A model is still running

A model is still running or downloading. Leaving will stop the current operation. Stay on this page to let it finish.

Skip to content
LocalAI.toolsBeta

Join the waitlist

Get an email when the new apps and model runs launch.

No account needed.

Qwen3-TTS

Generate local speech in a reference voice with Qwen3-TTS 0.6B Base. Supply target text, a clean PCM WAV recording and its exact transcript; returns complete 24kHz mono WAV audio.

Desktop inferencePrivate · Runs on your computer
About this model

Generate local speech in a reference voice with Qwen3-TTS 0.6B Base. Supply target text, a clean PCM WAV recording and its exact transcript; returns complete 24kHz mono WAV audio.

License
Apache-2.0
Model released
Jan 22, 2026
Source repository created
Jan 21, 2026
Model weights · FP32
2.5 GB

Best for

  • Permission-based local reference-voice speech synthesis
  • Synthetic-voice prototyping and reviewed audio production

Keep in mind

  • Voice similarity, pronunciation, prosody and words can drift; generated speech is not proof of identity or an authentic recording. Review output before use.
  • Use recordings you own or have permission to clone. Generated audio has no adapter-added watermark; do not use it to impersonate or deceive.
  • Noisy, clipped, overlapping or mis-transcribed references reduce quality. Clean single-speaker recordings around3–10seconds are recommended; PCM WAV only.
  • Base supports reference voice cloning, not description-based VoiceDesign. No streaming or arbitrary custom-speaker preset API is advertised.
  • Short English phrases with one synthetic reference passed independent ASR and coarse speaker-similarity gates. Natural speakers, other languages and broad perceptual/safety quality remain unverified.
  • Fixed seeds are not bit-stable across warm runs, devices or framework changes. Samples retain exact measured WAVs rather than reconstructed audio.
  • No fixed reference-duration or file-size cap: model context, audio-token budget, memory and the30-minute helper request timeout bound practical runs. Incomplete generation raises an error.
  • CPU FP32 and Mac Metal FP32/BF16 passed on the tested M3 Max. Metal FP16 failed nonfinite sampling and is unsupported; Windows and other Mac hardware/releases remain unverified.
  • Model weights and isolated app-owned runtime download only on request. System Python and user packages are not used.

Processed on your deviceYour input is processed locally in the desktop app. Model and runtime files download from external hosts.

Target platforms · verification required

  • macos · aarch64, x86_64
  • windows · x86_64

Recommended configuration

FP32 · cpu (verified platform only)

Sources & technical details
Original model
https://huggingface.co/Qwen/Qwen3-TTS-12Hz-0.6B-Base
Release verification
Verified
Release evidence
Official news explicitly releases Qwen3-TTS 0.6B/1.7B series; exact Base variant belongs to this initial release.
Source updated
Jan 29, 2026
Native model files
https://huggingface.co/Qwen/Qwen3-TTS-12Hz-0.6B-Base
Files updated
Jan 29, 2026
Artifact revision
5d83992436eae1d760afd27aff78a71d676296fc

Desktop app

Sample result

Qwen3 Voice Cloning — 0.6B Base

Run Qwen3-TTS on your computer

Download the desktop app to load this model, provide input and view results locally.

Public desktop downloads are coming soon.

Device requirements

macOS Apple SiliconCPU / FP32Tested16.0 GiBRAM estimate
macOS Apple SiliconMPS / FP32Tested16.0 GiBRAM estimate
macOS Apple SiliconMPS / BF16Tested16.0 GiBRAM estimate
Windows Intel / AMDCPU / FP32Unverified16.0 GiBRAM estimate
macOS Intel / AMDCPU / FP32Unverified16.0 GiBRAM estimate
If your device falls short

MemoryAdvisory only: close other apps, reduce image/tile/token/batch settings or choose a smaller verified model/precision. CPU can need more RAM than GPU.

GPUIf the selected GPU runtime fails, existing CPU fallback may recover when enabled, but CPU has its own memory/runtime needs.

PlatformUse a compatible OS/architecture/runtime or another model. These estimates do not add hardware detection or block runs.

Benchmark evidence

Saved desktop measurements apply to the recorded hardware and runtime. Fallback cases report the device that actually ran; they do not prove GPU support on the requested device.

Qwen3 Voice Cloning — 0.6B Base

46 configurations tested
Passed44
Failed2
TestedNov 1, 2026
Runtimeqwen-tts 0.1.1

Quality criteria

No explicit quality gate recorded. Review per-case evidence; full certification is pending.

PrecisionDeviceResultPlatformInference / loadTask metric
FP32mpsAcceleratedPassedModeratemacos · aarch644058 ms / 0 ms2.029 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch644098 ms / 0 ms2.049 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch644101 ms / 0 ms2.050 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch644236 ms / 0 ms2.118 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch644238 ms / 0 ms2.119 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch644272 ms / 0 ms2.136 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch644315 ms / 0 ms2.157 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch644349 ms / 0 ms2.175 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch644358 ms / 0 ms2.179 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch644368 ms / 0 ms2.184 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch644384 ms / 0 ms2.192 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch644473 ms / 0 ms2.237 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch644502 ms / 0 ms2.251 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch643781 ms / 0 ms2.363 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch645106 ms / 0 ms2.364 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch645236 ms / 1065 ms2.517 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch645249 ms / 1129 ms2.524 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch646297 ms / 992 ms2.539 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch646304 ms / 1002 ms2.542 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch643460 ms / 0 ms2.544 RTF
FP32mpsAcceleratedPassedModeratemacos · aarch645303 ms / 1067 ms2.549 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch646382 ms / 1005 ms2.573 RTF
FP32cpuPassedModeratemacos · aarch643914 ms / 0 ms1.578 RTF
FP32cpuPassedModeratemacos · aarch644084 ms / 641 ms1.595 RTF
FP32cpuPassedModeratemacos · aarch644097 ms / 612 ms1.600 RTF
FP32cpuPassedModeratemacos · aarch644184 ms / 0 ms1.634 RTF
FP32cpuPassedModeratemacos · aarch644185 ms / 0 ms1.635 RTF
FP32cpuPassedModeratemacos · aarch644210 ms / 600 ms1.645 RTF
FP32cpuPassedModeratemacos · aarch644256 ms / 0 ms1.663 RTF
FP32cpuPassedModeratemacos · aarch644264 ms / 0 ms1.666 RTF
FP32cpuPassedModeratemacos · aarch644264 ms / 0 ms1.666 RTF
FP32cpuPassedModeratemacos · aarch644275 ms / 0 ms1.670 RTF
FP32cpuPassedModeratemacos · aarch644320 ms / 0 ms1.688 RTF
FP32cpuPassedModeratemacos · aarch644328 ms / 0 ms1.691 RTF
FP32cpuPassedModeratemacos · aarch644329 ms / 0 ms1.691 RTF
FP32cpuPassedModeratemacos · aarch644330 ms / 0 ms1.691 RTF
FP32cpuPassedModeratemacos · aarch644336 ms / 0 ms1.694 RTF
FP32cpuPassedModeratemacos · aarch644376 ms / 0 ms1.709 RTF
FP32cpuPassedModeratemacos · aarch643159 ms / 0 ms1.717 RTF
FP32cpuPassedModeratemacos · aarch644407 ms / 816 ms1.721 RTF
FP32cpuPassedModeratemacos · aarch644281 ms / 0 ms1.726 RTF
FP32cpuPassedModeratemacos · aarch644456 ms / 815 ms1.741 RTF
FP32cpuPassedModeratemacos · aarch644468 ms / 811 ms1.745 RTF
FP32cpuPassedModeratemacos · aarch643362 ms / 0 ms1.827 RTF
FP32cpuFailed—Not recorded— / ——
FP16mpsFailed—Not recorded— / ——
Test setup & sample details
Quality gate
No explicit quality gate recorded. Review the per-case quality evidence; full certification is pending.
Tested
2026-11-01T20:36:24.746813+00:00
Environment
{ "hardware": "Apple M3 Max", "os": "macOS-27.2-arm64-arm-64bit-Mach-O", "python": "3.13.18", "runtime": "qwen-tts0.1.1/Transformers4.57.3/Torch2.10.0", "interpreterScope": "App-owned pinned CPython3.13.18 private probe venv; packaged default provision tested separately" }
Runtime
qwen-tts 0.1.1
Artifact revision
5d83992436eae1d760afd27aff78a71d676296fc
Fixture
{
  "file": "model-samples/desktop/qwen-voice/qwen-voice-reference.wav",
  "size": 140204,
  "sha1": "a116dbdfac0bda087bb6ebf5677cddb7d86b75fc",
  "sha256": "bac94d622d3bb0a32ece9b8446daf5f03257125ef8ffb480e59adfde18849105",
  "url": "https://assets.localai.tools/samples/v1/files/bac94d622d3bb0a32ece9b8446daf5f03257125ef8ffb480e59adfde18849105/qwen-voice-reference.wav"
}
Rerun when
artifact revision/checksum; qwen-tts/Transformers/Torch/audio dependency version or OS/device; sampling/generation/PCM preprocessing policy; reference/text fixtures/ASR/speaker evaluator/quality gates
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 815.4003329982515,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4461.386000002676,
      "inferenceMs": 4455.99483299884,
      "realTimeFactor": 1.740622981640172,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 5276.786333000928,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8120101690292358,
      "rms": 0.027086924761533737,
      "audioSeconds": 2.48,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.301415+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4282.595917000435,
      "inferenceMs": 4281.16354100348,
      "realTimeFactor": 1.726275621372371,
      "audioSeconds": 2.48,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 32,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.027086865156888962,
      "peak": 0.20531806349754333,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 4282.595917000435,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8178822994232178,
      "rms": 0.02787153609097004,
      "audioSeconds": 1.84,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.455867+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 3362.9900839951006,
      "inferenceMs": 3361.7826249974314,
      "realTimeFactor": 1.8270557744551257,
      "audioSeconds": 1.84,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 24,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 7,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.027871662750840187,
      "peak": 0.1949378103017807,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 3362.9900839951006,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4331.34487499774,
      "inferenceMs": 4330.075666999619,
      "realTimeFactor": 1.6914358074217262,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 4331.34487499774,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4329.6191670015105,
      "inferenceMs": 4328.177833995142,
      "realTimeFactor": 1.6906944664043522,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 4329.6191670015105,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 810.9440420012106,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4473.575333999179,
      "inferenceMs": 4468.289542004641,
      "realTimeFactor": 1.7454256023455628,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 5284.5193760003895,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4337.474790998385,
      "inferenceMs": 4336.174291995121,
      "realTimeFactor": 1.6938180828105942,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 4337.474790998385,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4330.215624999255,
      "inferenceMs": 4328.90083300299,
      "realTimeFactor": 1.690976887891793,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 4330.215624999255,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 815.9165830002166,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4412.183250002272,
      "inferenceMs": 4406.741040998895,
      "realTimeFactor": 1.7213832191401934,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 5228.099833002489,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4376.846207997005,
      "inferenceMs": 4375.552792000235,
      "realTimeFactor": 1.7092003093750918,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 4376.846207997005,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4321.321874995192,
      "inferenceMs": 4320.042540995928,
      "realTimeFactor": 1.6875166175765344,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 1,
      "totalMs": 4321.321874995192,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 600.1501669961726,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4215.354874999321,
      "inferenceMs": 4210.04725000239,
      "realTimeFactor": 1.6445497070321835,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4815.505041995493,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8120102286338806,
      "rms": 0.027086926624178886,
      "audioSeconds": 2.48,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.655681+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 3914.8853329970734,
      "inferenceMs": 3913.5087090035086,
      "realTimeFactor": 1.5780277052433502,
      "audioSeconds": 2.48,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 32,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.027086865156888962,
      "peak": 0.20531822741031647,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 3914.8853329970734,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8178822994232178,
      "rms": 0.02787153609097004,
      "audioSeconds": 1.84,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.455867+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 3160.008332997677,
      "inferenceMs": 3158.763416999136,
      "realTimeFactor": 1.7167192483690956,
      "audioSeconds": 1.84,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 24,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 7,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.027871662750840187,
      "peak": 0.1949378103017807,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 3160.008332997677,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4276.20300000126,
      "inferenceMs": 4274.77912500035,
      "realTimeFactor": 1.6698355957032618,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4276.20300000126,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4185.481874999823,
      "inferenceMs": 4184.041708002042,
      "realTimeFactor": 1.6343912921882975,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4185.481874999823,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 641.0416669968981,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4089.377083997533,
      "inferenceMs": 4084.0165829940815,
      "realTimeFactor": 1.595318977732063,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4730.418750994431,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4185.954291002417,
      "inferenceMs": 4184.64916600351,
      "realTimeFactor": 1.634628580470121,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4185.954291002417,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4265.326583001297,
      "inferenceMs": 4264.070500001253,
      "realTimeFactor": 1.6656525390629895,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4265.326583001297,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 611.7505829970469,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4102.229083997372,
      "inferenceMs": 4096.814875003474,
      "realTimeFactor": 1.6003183105482321,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4713.979666994419,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4265.843790999497,
      "inferenceMs": 4264.318749999802,
      "realTimeFactor": 1.6657495117186727,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4265.843790999497,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7734778523445129,
      "rms": 0.02550254762172699,
      "audioSeconds": 2.56,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.142061+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4257.65183299518,
      "inferenceMs": 4256.361375002598,
      "realTimeFactor": 1.6626411621103898,
      "audioSeconds": 2.56,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 33,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.025502532720565796,
      "peak": 0.1848255842924118,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4257.65183299518,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7602836489677429,
      "rms": 0.037735648453235626,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.835374+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1128.7702089975937,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5254.625833993487,
      "inferenceMs": 5248.899375001201,
      "realTimeFactor": 2.5235093149044237,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.0377357080578804,
      "peak": 0.23305727541446686,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5896912896,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6383.39604299108,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8448126316070557,
      "rms": 0.028364835307002068,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.043510+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4316.204209004354,
      "inferenceMs": 4314.814334000403,
      "realTimeFactor": 2.1574071670002013,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.028364794328808784,
      "peak": 0.18572819232940674,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5906759680,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4316.204209004354,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7797691226005554,
      "rms": 0.017842229455709457,
      "audioSeconds": 1.6,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.229472+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 3782.316291995812,
      "inferenceMs": 3780.989500002761,
      "realTimeFactor": 2.3631184375017256,
      "audioSeconds": 1.6,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 21,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 7,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.01784219965338707,
      "peak": 0.1772211790084839,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5908774912,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 3782.316291995812,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7940500974655151,
      "rms": 0.03727245330810547,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.413467+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4059.5561250011087,
      "inferenceMs": 4058.261084006517,
      "realTimeFactor": 2.0291305420032586,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03727246820926666,
      "peak": 0.2534540593624115,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5908774912,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4059.5561250011087,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7940500974655151,
      "rms": 0.03727245330810547,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.413467+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4237.332542004879,
      "inferenceMs": 4236.115708001307,
      "realTimeFactor": 2.1180578540006536,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03727246820926666,
      "peak": 0.2534540593624115,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5908774912,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4237.332542004879,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7602836489677429,
      "rms": 0.037735648453235626,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.835374+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1066.8164999951841,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5307.941624996602,
      "inferenceMs": 5302.627458004281,
      "realTimeFactor": 2.5493401240405196,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.0377357080578804,
      "peak": 0.23305727541446686,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5884329984,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6374.758124991786,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7940500974655151,
      "rms": 0.03727245330810547,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.413467+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4273.1971250032075,
      "inferenceMs": 4271.7599169991445,
      "realTimeFactor": 2.1358799584995722,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03727246820926666,
      "peak": 0.2534540593624115,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5906759680,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4273.1971250032075,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7940500974655151,
      "rms": 0.03727245330810547,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.413467+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4099.572958002682,
      "inferenceMs": 4098.349375002726,
      "realTimeFactor": 2.0491746875013632,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03727246820926666,
      "peak": 0.2534540593624115,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5906759680,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4099.572958002682,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7602836489677429,
      "rms": 0.037735648453235626,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:25.835374+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1065.4793749999953,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5241.52483400394,
      "inferenceMs": 5236.036791997321,
      "realTimeFactor": 2.517325380767943,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.0377357080578804,
      "peak": 0.23305727541446686,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5896912896,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 6307.004209003935,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7940500974655151,
      "rms": 0.03727245330810547,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.413467+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4474.331208002695,
      "inferenceMs": 4473.039666998375,
      "realTimeFactor": 2.2365198334991874,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03727246820926666,
      "peak": 0.2534540593624115,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5906759680,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4474.331208002695,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7940500974655151,
      "rms": 0.03727245330810547,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.413467+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4102.12679200049,
      "inferenceMs": 4100.884500003303,
      "realTimeFactor": 2.0504422500016517,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03727246820926666,
      "peak": 0.2534540593624115,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5906759680,
      "precision": "fp32",
      "cpuThreads": 4,
      "totalMs": 4102.12679200049,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.842760443687439,
      "rms": 0.027899807319045067,
      "audioSeconds": 2.48,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.608958+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1001.6120000000228,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 6309.626958995068,
      "inferenceMs": 6303.733333996206,
      "realTimeFactor": 2.5418279572565345,
      "audioSeconds": 2.48,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 32,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.027899807319045067,
      "peak": 0.244140625,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3225436160,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 7311.238958995091,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7947965860366821,
      "rms": 0.028488820418715477,
      "audioSeconds": 2.16,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.810156+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 5107.647458004067,
      "inferenceMs": 5106.339707999723,
      "realTimeFactor": 2.364046161110983,
      "audioSeconds": 2.16,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 28,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.02848881296813488,
      "peak": 0.1767578125,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3226894336,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 5107.647458004067,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.777942419052124,
      "rms": 0.02627979964017868,
      "audioSeconds": 1.36,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:27.042914+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 3460.809167001571,
      "inferenceMs": 3459.5860410045134,
      "realTimeFactor": 2.543813265444495,
      "audioSeconds": 1.36,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 18,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 7,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.026279792189598083,
      "peak": 0.2119140625,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3235119104,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 3460.809167001571,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7772647738456726,
      "rms": 0.02264554239809513,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:27.153945+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 4503.383709001355,
      "inferenceMs": 4502.139291995263,
      "realTimeFactor": 2.2510696459976316,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.022645534947514534,
      "peak": 0.1376953125,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3236544512,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 4503.383709001355,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7772647738456726,
      "rms": 0.02264554239809513,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:27.153945+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 4350.490082993929,
      "inferenceMs": 4349.202708006487,
      "realTimeFactor": 2.1746013540032436,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.022645534947514534,
      "peak": 0.1376953125,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3236544512,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 4350.490082993929,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.842760443687439,
      "rms": 0.027899807319045067,
      "audioSeconds": 2.48,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.608958+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1005.4294580040732,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 6387.319082998147,
      "inferenceMs": 6381.831332997535,
      "realTimeFactor": 2.573319085886103,
      "audioSeconds": 2.48,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 32,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.027899807319045067,
      "peak": 0.244140625,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3225436160,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 7392.74854100222,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7772647738456726,
      "rms": 0.02264554239809513,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:27.153945+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 4385.306582997146,
      "inferenceMs": 4384.053708003194,
      "realTimeFactor": 2.192026854001597,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.022645534947514534,
      "peak": 0.1376953125,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3233153024,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 4385.306582997146,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7772647738456726,
      "rms": 0.02264554239809513,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:27.153945+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 4358.962917001918,
      "inferenceMs": 4357.726374997583,
      "realTimeFactor": 2.1788631874987914,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.022645534947514534,
      "peak": 0.1376953125,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3233153024,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 4358.962917001918,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.842760443687439,
      "rms": 0.027899807319045067,
      "audioSeconds": 2.48,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:26.608958+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 992.2057090007002,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 6302.59166599717,
      "inferenceMs": 6297.137292000116,
      "realTimeFactor": 2.5391682629032726,
      "audioSeconds": 2.48,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 32,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.027899807319045067,
      "peak": 0.244140625,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3225436160,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 7294.79737499787,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7772647738456726,
      "rms": 0.02264554239809513,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:27.153945+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 4369.463250004628,
      "inferenceMs": 4367.941541997425,
      "realTimeFactor": 2.1839707709987124,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.022645534947514534,
      "peak": 0.1376953125,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3233153024,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 4369.463250004628,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7772647738456726,
      "rms": 0.02264554239809513,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-01T20:42:27.153945+00:00",
      "structuralPassed": true,
      "identicalAudioQualityReused": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 0,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 4239.29729199881,
      "inferenceMs": 4238.061125004606,
      "realTimeFactor": 2.119030562502303,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.022645534947514534,
      "peak": 0.1376953125,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 3233153024,
      "precision": "bf16",
      "cpuThreads": 4,
      "totalMs": 4239.29729199881,
      "artifactRevision": "5d83992436eae1d760afd27aff78a71d676296fc",
      "fallback": false,
      "grade": "moderate"
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": "Not embedded",
    "failure": "Generation reached its token budget without EOS; increase maxNewTokens"
  },
  {
    "precision": "fp16",
    "device": "mps",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": "Not embedded",
    "failure": "probability tensor contains either `inf`, `nan` or element < 0"
  }
]

Qwen3-TTS 1.7B · Base — Voice Cloning

28 configurations tested
Passed24
Failed4
TestedNov 6, 2026
Runtimemlx-audio 0.6.3

Quality criteria

{ "maximumWordErrorRate": 0.25, "minimumSpeakerCosine": 0.6, "minimumRms": 0.005, "maximumClippedFraction": 0.02, "eosRequired": true, "maximumRealTimeFactor": 3, "maximumRequestMs": 45000 }

PrecisionDeviceResultPlatformInference / loadTask metric
BF16mpsAcceleratedPassedModeratemacos · aarch64985 ms / 321 ms0.513 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641270 ms / 321 ms0.661 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch64904 ms / 321 ms0.664 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641325 ms / 328 ms0.690 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641331 ms / 328 ms0.693 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch64952 ms / 328 ms0.700 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641359 ms / 324 ms0.708 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641448 ms / 328 ms0.724 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641479 ms / 321 ms0.740 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641430 ms / 324 ms0.745 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641592 ms / 324 ms0.796 RTF
BF16mpsAcceleratedPassedModeratemacos · aarch641210 ms / 324 ms0.889 RTF
FP32cpuPassedModeratemacos · aarch645357 ms / 932 ms2.309 RTF
FP32cpuPassedModeratemacos · aarch645375 ms / 967 ms2.317 RTF
FP32cpuPassedModeratemacos · aarch645379 ms / 967 ms2.318 RTF
FP32cpuPassedModeratemacos · aarch645447 ms / 911 ms2.348 RTF
FP32cpuPassedModeratemacos · aarch645476 ms / 911 ms2.360 RTF
FP32cpuPassedModeratemacos · aarch644919 ms / 911 ms2.365 RTF
FP32cpuPassedModeratemacos · aarch644980 ms / 967 ms2.394 RTF
FP32cpuPassedModeratemacos · aarch645578 ms / 932 ms2.404 RTF
FP32cpuPassedModeratemacos · aarch645100 ms / 932 ms2.452 RTF
FP32cpuPassedModeratemacos · aarch643923 ms / 967 ms2.452 RTF
FP32cpuPassedModeratemacos · aarch643934 ms / 932 ms2.459 RTF
FP32cpuPassedModeratemacos · aarch644025 ms / 911 ms2.516 RTF
BF16mpsFailed—macos · aarch649878 ms / 1455 ms4.749 RTF
FP32mpsFailed—macos · aarch648265 ms / 1666 ms3.973 RTF
FP16mpsFailed—macos · aarch649452 ms / 1416 ms4.220 RTF
FP32mpsFailed—macos · aarch649877 ms / 1640 ms4.749 RTF
Test setup & sample details
Quality gate
{ "maximumWordErrorRate": 0.25, "minimumSpeakerCosine": 0.6, "minimumRms": 0.005, "maximumClippedFraction": 0.02, "eosRequired": true, "maximumRealTimeFactor": 3, "maximumRequestMs": 45000 }
Tested
2026-11-06T14:34:13.031046+00:00
Environment
{ "os": "macos", "arch": "aarch64", "hardware": "Apple M3 Max", "osVersion": "27.2", "runtime": "App-owned CPython3.13.18; CPU qwen-tts0.1.1/Transformers4.57.3/Torch2.10.0; Metal mlx-audio0.6.3/MLX0.32.4/Transformers5.22.1; original identical checkpoint files" }
Runtime
mlx-audio 0.6.3
Artifact revision
fd4b254389122332181a7c3db7f27e918eec64e3
Fixture
{
  "file": "model-samples/desktop/qwen3-tts-family/base-morning-input.json",
  "url": "https://assets.localai.tools/samples/v1/files/9eeb095d7a37292dd47c20668da49bbe84dc00547169f6ee2cd74a2fe95def0f/base-morning-input.json",
  "size": 696,
  "sha1": "0d74398daf9d5f2607c94d8811b0abeb9fb5aa09",
  "sha256": "9eeb095d7a37292dd47c20668da49bbe84dc00547169f6ee2cd74a2fe95def0f"
}
Rerun when
artifact revision/checksum; qwen-tts/Transformers/Torch/audio dependency version or OS/device; sampling/generation/PCM preprocessing policy; reference/text fixtures/ASR/speaker evaluator/quality gates
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.789526104927063,
      "rms": 0.0341080017387867,
      "audioSeconds": 2.32,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:41.481898+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 911.4559160007047,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5481.678750000356,
      "inferenceMs": 5476.0258339993015,
      "realTimeFactor": 2.3603559629307336,
      "audioSeconds": 2.32,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 30,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03410801663994789,
      "peak": 0.21620424091815948,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 6408.511249001094,
      "audioSecondsPerSecond": 0.4236649114391844,
      "grade": "moderate",
      "residentMemoryBytes": 10300194816,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8278511762619019,
      "rms": 0.02585652843117714,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:41.595348+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 911.4559160007047,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4920.6764999998995,
      "inferenceMs": 4919.072791000872,
      "realTimeFactor": 2.364938841827342,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.02585645206272602,
      "peak": 0.15182772278785706,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 5846.877749001578,
      "audioSecondsPerSecond": 0.4228439155861297,
      "grade": "moderate",
      "residentMemoryBytes": 10322198528,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8441740274429321,
      "rms": 0.023547956719994545,
      "audioSeconds": 1.6,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:41.689538+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 911.4559160007047,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4026.8038340000203,
      "inferenceMs": 4025.4451250002603,
      "realTimeFactor": 2.5159032031251627,
      "audioSeconds": 1.6,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 21,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 7,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.023547951132059097,
      "peak": 0.17534145712852478,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 4951.911291002034,
      "audioSecondsPerSecond": 0.3974715715445001,
      "grade": "moderate",
      "residentMemoryBytes": 10391289856,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.789526104927063,
      "rms": 0.0341080017387867,
      "audioSeconds": 2.32,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:41.806702+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 911.4559160007047,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5448.070958000244,
      "inferenceMs": 5446.576250000362,
      "realTimeFactor": 2.347662176724294,
      "audioSeconds": 2.32,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 30,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03410801663994789,
      "peak": 0.21620424091815948,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 6374.739458000477,
      "audioSecondsPerSecond": 0.42595566343165503,
      "grade": "moderate",
      "residentMemoryBytes": 10391552000,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.789526104927063,
      "rms": 0.0341080017387867,
      "audioSeconds": 2.32,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:41.920566+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 967.3893749986746,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5384.659041999839,
      "inferenceMs": 5378.870582999298,
      "realTimeFactor": 2.318478699568663,
      "audioSeconds": 2.32,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 30,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03410801663994789,
      "peak": 0.21620424091815948,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 6368.2484999990265,
      "audioSecondsPerSecond": 0.43131731172947296,
      "grade": "moderate",
      "residentMemoryBytes": 10284122112,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8278511762619019,
      "rms": 0.02585652843117714,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.035831+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 967.3893749986746,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 4981.242500000008,
      "inferenceMs": 4979.718457998388,
      "realTimeFactor": 2.3940954124992246,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.02585645206272602,
      "peak": 0.15182772278785706,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 5963.828541000112,
      "audioSecondsPerSecond": 0.41769429688522236,
      "grade": "moderate",
      "residentMemoryBytes": 10302275584,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8441740274429321,
      "rms": 0.023547956719994545,
      "audioSeconds": 1.6,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.130249+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 967.3893749986746,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 3924.6526250008174,
      "inferenceMs": 3923.2425419995707,
      "realTimeFactor": 2.4520265887497317,
      "audioSeconds": 1.6,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 21,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 7,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.023547951132059097,
      "peak": 0.17534145712852478,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 4906.3146249973215,
      "audioSecondsPerSecond": 0.4078259202360003,
      "grade": "moderate",
      "residentMemoryBytes": 10370531328,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.789526104927063,
      "rms": 0.0341080017387867,
      "audioSeconds": 2.32,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.241483+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 967.3893749986746,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5376.775749999069,
      "inferenceMs": 5375.253582998994,
      "realTimeFactor": 2.316919647844394,
      "audioSeconds": 2.32,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 30,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03410801663994789,
      "peak": 0.21620424091815948,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 6360.661374999836,
      "audioSecondsPerSecond": 0.43160754449571687,
      "grade": "moderate",
      "residentMemoryBytes": 10370531328,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.789526104927063,
      "rms": 0.0341080017387867,
      "audioSeconds": 2.32,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.358944+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 931.5446250002424,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5363.031374999991,
      "inferenceMs": 5357.397916000991,
      "realTimeFactor": 2.3092232396555996,
      "audioSeconds": 2.32,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 30,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03410801663994789,
      "peak": 0.21620424091815948,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 6310.130333000416,
      "audioSecondsPerSecond": 0.4330460489169255,
      "grade": "moderate",
      "residentMemoryBytes": 10281025536,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8278511762619019,
      "rms": 0.02585652843117714,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.477433+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 931.5446250002424,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5101.22945799958,
      "inferenceMs": 5099.744708999424,
      "realTimeFactor": 2.4518003408651077,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.02585645206272602,
      "peak": 0.15182772278785706,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 6047.717665998789,
      "audioSecondsPerSecond": 0.4078635537048478,
      "grade": "moderate",
      "residentMemoryBytes": 10314907648,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8441740274429321,
      "rms": 0.023547956719994545,
      "audioSeconds": 1.6,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.571870+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 931.5446250002424,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 3935.7535830004053,
      "inferenceMs": 3934.4094170010067,
      "realTimeFactor": 2.459005885625629,
      "audioSeconds": 1.6,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 21,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 7,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.023547951132059097,
      "peak": 0.17534145712852478,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 4880.585540999164,
      "audioSecondsPerSecond": 0.4066684044335162,
      "grade": "moderate",
      "residentMemoryBytes": 10371399680,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "cpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.789526104927063,
      "rms": 0.0341080017387867,
      "audioSeconds": 2.32,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.679604+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 931.5446250002424,
      "placement": {
        "talker": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "cpu",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "cpu",
          "dtype": "torch.float32"
        }
      },
      "device": "cpu",
      "provider": "PyTorch CPU",
      "acceleration": "cpu",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 5579.605625000113,
      "inferenceMs": 5578.17950000026,
      "realTimeFactor": 2.4043877155173536,
      "audioSeconds": 2.32,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 30,
      "generationPlacement": "cpu",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03410801663994789,
      "peak": 0.21620424091815948,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 6528.361417000269,
      "audioSecondsPerSecond": 0.41590630061293143,
      "grade": "moderate",
      "residentMemoryBytes": 10376052736,
      "supportsStreaming": false
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7780048847198486,
      "rms": 0.027339505031704903,
      "audioSeconds": 1.92,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.780085+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 321.1342919985327,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 994.7397089999868,
      "inferenceMs": 985.212916000819,
      "realTimeFactor": 0.5131317270837599,
      "audioSeconds": 1.92,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 25,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.027339495718479156,
      "peak": 0.2151423990726471,
      "clippedFraction": 0,
      "chunks": 4,
      "firstAudioMs": 308.2117499998276,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616285216,
      "peakMemoryBytes": 5270309029,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1331.258041998808,
      "audioSecondsPerSecond": 1.948817325491096,
      "grade": "moderate",
      "residentMemoryBytes": 4923179008
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8384092450141907,
      "rms": 0.032424572855234146,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:42.881021+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 321.1342919985327,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1485.9613749995333,
      "inferenceMs": 1479.1412079994188,
      "realTimeFactor": 0.7395706039997094,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.03242456167936325,
      "peak": 0.24000541865825653,
      "clippedFraction": 0,
      "chunks": 5,
      "firstAudioMs": 260.39612499880604,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616293412,
      "peakMemoryBytes": 5270452355,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1823.8018329993793,
      "audioSecondsPerSecond": 1.352135948335222,
      "grade": "moderate",
      "residentMemoryBytes": 4925915136
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8221558928489685,
      "rms": 0.02968289703130722,
      "audioSeconds": 1.36,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.041286+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 321.1342919985327,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 911.6160409994336,
      "inferenceMs": 903.5887499994715,
      "realTimeFactor": 0.6644034926466702,
      "audioSeconds": 1.36,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 18,
      "maxNewTokens": 512,
      "promptTokens": 7,
      "rms": 0.029682910069823265,
      "peak": 0.19976216554641724,
      "clippedFraction": 0,
      "chunks": 3,
      "firstAudioMs": 297.55829100031406,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616236036,
      "peakMemoryBytes": 5270452355,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1248.556500999257,
      "audioSecondsPerSecond": 1.505109487032453,
      "grade": "moderate",
      "residentMemoryBytes": 4925915136
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7780048847198486,
      "rms": 0.027339505031704903,
      "audioSeconds": 1.92,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.141763+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 321.1342919985327,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1277.9830829986167,
      "inferenceMs": 1270.046542000273,
      "realTimeFactor": 0.6614825739584755,
      "audioSeconds": 1.92,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 25,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.027339495718479156,
      "peak": 0.2151423990726471,
      "clippedFraction": 0,
      "chunks": 4,
      "firstAudioMs": 346.3130420004745,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616285216,
      "peakMemoryBytes": 5270554775,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1617.129999998724,
      "audioSecondsPerSecond": 1.5117556219444337,
      "grade": "moderate",
      "residentMemoryBytes": 4925849600
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7780048847198486,
      "rms": 0.027339505031704903,
      "audioSeconds": 1.92,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.239020+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 327.94654200006335,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1334.0037079997273,
      "inferenceMs": 1324.8260420004954,
      "realTimeFactor": 0.6900135635419247,
      "audioSeconds": 1.92,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 25,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.027339495718479156,
      "peak": 0.2151423990726471,
      "clippedFraction": 0,
      "chunks": 4,
      "firstAudioMs": 316.721499999403,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616283168,
      "peakMemoryBytes": 5270651071,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1679.8097509999934,
      "audioSecondsPerSecond": 1.4492468740279203,
      "grade": "moderate",
      "residentMemoryBytes": 4921933824
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8384092450141907,
      "rms": 0.032424572855234146,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.349853+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 327.94654200006335,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1457.8750410000794,
      "inferenceMs": 1448.089625000648,
      "realTimeFactor": 0.724044812500324,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.03242456167936325,
      "peak": 0.24000541865825653,
      "clippedFraction": 0,
      "chunks": 5,
      "firstAudioMs": 406.85170799952175,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616291364,
      "peakMemoryBytes": 5270651071,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1809.2211249986576,
      "audioSecondsPerSecond": 1.381129983580336,
      "grade": "moderate",
      "residentMemoryBytes": 4924456960
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8221558928489685,
      "rms": 0.02968289703130722,
      "audioSeconds": 1.36,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.443378+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 327.94654200006335,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 962.9559579989291,
      "inferenceMs": 952.2022090004612,
      "realTimeFactor": 0.7001486830885744,
      "audioSeconds": 1.36,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 18,
      "maxNewTokens": 512,
      "promptTokens": 7,
      "rms": 0.029682910069823265,
      "peak": 0.19976216554641724,
      "clippedFraction": 0,
      "chunks": 3,
      "firstAudioMs": 378.56720899981156,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616233988,
      "peakMemoryBytes": 5270659369,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1311.5315009999904,
      "audioSecondsPerSecond": 1.4282680581340064,
      "grade": "moderate",
      "residentMemoryBytes": 4925259776
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7780048847198486,
      "rms": 0.027339505031704903,
      "audioSeconds": 1.92,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.543815+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 327.94654200006335,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1340.5042909998883,
      "inferenceMs": 1331.169167000553,
      "realTimeFactor": 0.6933172744794547,
      "audioSeconds": 1.92,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 25,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.027339495718479156,
      "peak": 0.2151423990726471,
      "clippedFraction": 0,
      "chunks": 4,
      "firstAudioMs": 372.81112500022573,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616283168,
      "peakMemoryBytes": 5270659369,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1689.930959000776,
      "audioSecondsPerSecond": 1.4423410995360009,
      "grade": "moderate",
      "residentMemoryBytes": 4925325312
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7780048847198486,
      "rms": 0.027339505031704903,
      "audioSeconds": 1.92,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.640601+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 323.8117079999938,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1439.2289580009674,
      "inferenceMs": 1430.0270409985387,
      "realTimeFactor": 0.7448057505200723,
      "audioSeconds": 1.92,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 25,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.027339495718479156,
      "peak": 0.2151423990726471,
      "clippedFraction": 0,
      "chunks": 4,
      "firstAudioMs": 314.38470799912466,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616283168,
      "peakMemoryBytes": 5270118757,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1791.9998749985098,
      "audioSecondsPerSecond": 1.3426319537701399,
      "grade": "moderate",
      "residentMemoryBytes": 4921425920
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Please save the blue folder on your computer.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8384092450141907,
      "rms": 0.032424572855234146,
      "audioSeconds": 2,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.739460+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 323.8117079999938,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1610.4837080001744,
      "inferenceMs": 1592.0074160003423,
      "realTimeFactor": 0.7960037080001712,
      "audioSeconds": 2,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 26,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.03242456167936325,
      "peak": 0.24000541865825653,
      "clippedFraction": 0,
      "chunks": 5,
      "firstAudioMs": 355.69329100144387,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616291364,
      "peakMemoryBytes": 5270118757,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1981.167499001458,
      "audioSecondsPerSecond": 1.2562755549372202,
      "grade": "moderate",
      "residentMemoryBytes": 4924194816
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Your code is four two seven.",
      "wordErrorRate": 0,
      "speakerCosine": 0.8221558928489685,
      "rms": 0.02968289703130722,
      "audioSeconds": 1.36,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.832785+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 323.8117079999938,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1226.9241249996412,
      "inferenceMs": 1209.5263749997684,
      "realTimeFactor": 0.8893576286763003,
      "audioSeconds": 1.36,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 18,
      "maxNewTokens": 512,
      "promptTokens": 7,
      "rms": 0.029682910069823265,
      "peak": 0.19976216554641724,
      "clippedFraction": 0,
      "chunks": 3,
      "firstAudioMs": 357.00783299944305,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616233988,
      "peakMemoryBytes": 5270118757,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1583.7695829995937,
      "audioSecondsPerSecond": 1.1244070638809016,
      "grade": "moderate",
      "residentMemoryBytes": 4924325888
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "pass",
    "quality": {
      "passed": true,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7780048847198486,
      "rms": 0.027339505031704903,
      "audioSeconds": 1.92,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:43:43.940659+00:00",
      "structuralPassed": true,
      "performancePassed": true
    },
    "runtime": {
      "engine": "mlx-audio",
      "version": "0.6.3",
      "mlxVersion": "0.32.4",
      "transformersVersion": "5.22.1",
      "loadMs": 323.8117079999938,
      "device": "mps",
      "provider": "MLX Metal",
      "acceleration": "partial",
      "parameterDtype": "Original mixed BF16 talker / FP32 codec",
      "parameterDtypes": {
        "mlx.core.bfloat16": 480,
        "mlx.core.float32": 464
      },
      "placement": {
        "talker": "MLX default GPU",
        "speechTokenizer": "MLX default GPU",
        "speakerEncoder": "MLX default GPU"
      },
      "accelerationDetails": "Neural tensor evaluation on explicit MLX GPU; CPU tokenizer, WAV/resampling and serialization; no CPU device fallback",
      "supportsStreaming": true,
      "strictCodecLoading": true,
      "restoredOriginalCodebookStateFlags": 32,
      "durationMs": 1371.4757090001513,
      "inferenceMs": 1358.6566670001048,
      "realTimeFactor": 0.7076336807292213,
      "audioSeconds": 1.92,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 25,
      "maxNewTokens": 512,
      "promptTokens": 9,
      "rms": 0.027339495718479156,
      "peak": 0.2151423990726471,
      "clippedFraction": 0,
      "chunks": 4,
      "firstAudioMs": 341.61999999923864,
      "generationPlacement": "MLX GPU",
      "activeMemoryBytes": 4616283168,
      "peakMemoryBytes": 5270208667,
      "sampling": "Seeded .9/50/1.0/1.05; main/subtalker defaults from pinned MLX implementation",
      "taskMode": "base",
      "instruct": "",
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 1720.253125000454,
      "audioSecondsPerSecond": 1.413160547939851,
      "grade": "moderate",
      "residentMemoryBytes": 4925014016
    }
  },
  {
    "precision": "bf16",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.6840697526931763,
      "rms": 0.03587322682142258,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:28:58.657745+00:00",
      "structuralPassed": true,
      "performancePassed": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1455.1348330005567,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.bfloat16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.bfloat16",
      "durationMs": 9883.599125001638,
      "inferenceMs": 9877.612415999465,
      "realTimeFactor": 4.748852123076666,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03587322682142258,
      "peak": 0.25390625,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5278220288,
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "bf16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 11355.255791000673,
      "audioSecondsPerSecond": 0.21057720351842085,
      "grade": "slow",
      "residentMemoryBytes": 1243037696,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.781967282295227,
      "rms": 0.03563639521598816,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:28:58.835842+00:00",
      "structuralPassed": true,
      "performancePassed": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1665.672707998965,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 8271.273124999425,
      "inferenceMs": 8264.709250001033,
      "realTimeFactor": 3.973417908654343,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.035636402666568756,
      "peak": 0.22093716263771057,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 9980067840,
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 9951.718500000425,
      "audioSecondsPerSecond": 0.2516724953149126,
      "grade": "slow",
      "residentMemoryBytes": 1266745344,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp16",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7855304479598999,
      "rms": 0.03908950462937355,
      "audioSeconds": 2.24,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:28:59.027253+00:00",
      "structuralPassed": true,
      "performancePassed": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1415.945582999484,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float16"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float16"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float16"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float16",
      "durationMs": 9458.287708999706,
      "inferenceMs": 9451.883959000043,
      "realTimeFactor": 4.2195910531250185,
      "audioSeconds": 2.24,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 29,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03908966854214668,
      "peak": 0.2227783203125,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 5278351360,
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp16",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 10889.790207998885,
      "audioSecondsPerSecond": 0.2369897905768386,
      "grade": "slow",
      "residentMemoryBytes": 1254129664,
      "supportsStreaming": false
    }
  },
  {
    "precision": "fp32",
    "device": "mps",
    "status": "fail",
    "quality": {
      "passed": false,
      "transcript": "Good morning. The train leaves at nine.",
      "wordErrorRate": 0,
      "speakerCosine": 0.7745249271392822,
      "rms": 0.033392179757356644,
      "audioSeconds": 2.08,
      "repetitive": false,
      "gates": {
        "maximumWordErrorRate": 0.25,
        "minimumSpeakerCosine": 0.6,
        "minimumRms": 0.005,
        "maximumClippedFraction": 0.02,
        "durationSeconds": [
          0.5,
          20
        ],
        "eosRequired": true,
        "repetitionRejected": true,
        "scope": "Short English synthetic-reference voice cloning; ASR intelligibility and independent Resemblyzer voice similarity, not human identity/safety or broad multilingual/perceptual certification"
      },
      "evaluator": "Independent Qwen3-ASR0.6B MLXQ4 + Resemblyzer0.1.4 CPU; coarse synthetic-reference similarity, not biometric identity or safety certification",
      "evaluatedAt": "2026-11-06T14:29:03.476131+00:00",
      "structuralPassed": true,
      "performancePassed": false
    },
    "runtime": {
      "engine": "qwen-tts",
      "version": "0.1.1",
      "transformersVersion": "4.57.3",
      "torchVersion": "2.10.0",
      "loadMs": 1639.9747089999437,
      "placement": {
        "talker": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speakerEncoder": {
          "device": "mps:0",
          "dtype": "torch.float32"
        },
        "speechTokenizer": {
          "device": "mps:0",
          "dtype": "torch.float32"
        }
      },
      "device": "mps",
      "provider": "Metal/MPS",
      "acceleration": "partial",
      "accelerationDetails": "Talker, reference speaker encoder and speech tokenizer parameters on selected device; CPU WAV decoding/resampling, text preparation and encoding remain",
      "parameterDtype": "torch.float32",
      "durationMs": 9883.556791000956,
      "inferenceMs": 9877.3408750003,
      "realTimeFactor": 4.748721574519375,
      "audioSeconds": 2.08,
      "seed": 42,
      "eosReached": true,
      "generatedTokens": 27,
      "generationPlacement": "mps:0",
      "referenceSeconds": 2.92,
      "referenceSampleRate": 24000,
      "language": "English",
      "promptTokens": 9,
      "referenceTextTokens": 10,
      "maxNewTokens": 512,
      "rms": 0.03339218720793724,
      "peak": 0.22585250437259674,
      "clippedFraction": 0,
      "sampling": "Official seeded main/subtalker sampling: temperature0.9, topK50, topP1.0, repetitionPenalty1.05",
      "driverAllocatedMemoryBytes": 10021355520,
      "artifactRevision": "fd4b254389122332181a7c3db7f27e918eec64e3",
      "precision": "fp32",
      "cpuThreads": 4,
      "fallback": false,
      "totalMs": 11541.743042000235,
      "audioSecondsPerSecond": 0.21058299256073176,
      "grade": "slow",
      "residentMemoryBytes": 1344618496,
      "supportsStreaming": false
    }
  }
]

Something not quite right?

Tell us what happened with Qwen3-TTS. Loading, results, or the page itself—we’re here for all of it.

Attachments Optional

Up to 3 files · 50 MB each

Browser, OS, CPU/GPU, memory, screen, storage, and network details are included automatically. Model inputs, results, and browsing history are not collected.