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.

SpeechT5

English text in. Speech out. Generated on your device.

Browser inferencePowered by WebAI · Private · Unlimited runs
About this model

Read English text aloud using a speaker embedding to condition the voice. Returns 16 kHz mono audio; an optional speaker-embedding file must contain exactly 512 float32 values.

Created by
Microsoft
License
MIT
Model released
Unknown
Source repository created
Feb 2, 2023
Model weights · FP32
642.7 MB

Best for

  • Local English narration and accessibility prototypes.
  • Generating short, user-reviewed voice prompts in a browser.
  • Researching speaker-conditioned text-to-speech with supplied embeddings.

Keep in mind

  • The source model is trained for English; other languages and code-switching are unsupported.
  • Names, acronyms, numbers, punctuation, and uncommon words may be pronounced incorrectly.
  • Speaker embeddings influence voice characteristics but do not guarantee faithful identity reproduction.
  • Synthetic audio may contain omissions, repetitions, artifacts, or misleading prosody and must not be used for impersonation, deception, or safety-critical communication.

Processed on your deviceInference runs in this browser. Input/output sharing requires consent for each run. Basic run logs include device information and approximate connection location. Model and runtime files download from external hosts. Saved downloads may be cleared by your browser.

Sources & technical details
Original model
microsoft/speecht5_tts
Release verification
Unknown
Release evidence
Exact original checkpoint public release date has not been verified. Repository creation, training, paper, artifact conversion and update dates are not release evidence.
Source updated
Nov 8, 2023
Browser model files
Xenova/speecht5_tts
Files updated
Aug 27, 2025
Artifact revision
1723781b8ce2d02f0400c8337be04ae8ee3d6d56

Input

TEXT → AUDIO
55 characters · 1 clips

Sample result

SpeechT5 — English Text to Speech

Blank lines create separate clips. Keep each clip short; SpeechT5 supports up to 600 tokenizer positions.

Uses the worker’s pinned reference speaker embedding.

Not benchmarkedEstimated runtime

Output

Ready to give your words a voice

Enter text, choose a voice when available, and generate speech. Listen here or download a WAV file.

Benchmark evidence

Saved worker evidence is separate from benchmarks run on your device. It does not guarantee speed or quality on every browser.

SpeechT5 — English Text to Speech

16 configurations tested
Passed5
Failed11
TestedOct 23, 2026
RuntimeTransformers.js 3.7.6

Quality criteria

Finite, non-clipped, non-silent 16 kHz audio of plausible duration whose normalized 64-bin amplitude envelope has dynamic-time-warping distance at most 0.45 from the checked-in stochastic FP32/WASM reference.

PrecisionDeviceResult
FP32CPU · WASMPassedSlow
Q8CPU · WASMPassedSlow
FP16CPU · WASMPassedVery slow
Q4CPU · WASMPassedVery slow
Q4F16CPU · WASMPassedVery slow
FP32GPU · WebGPUFailed—
FP16GPU · WebGPUFailed—
INT8CPU · WASMFailed—
INT8GPU · WebGPUFailed—
UINT8CPU · WASMFailed—
UINT8GPU · WebGPUFailed—
Q8GPU · WebGPUFailed—
Q4GPU · WebGPUFailed—
Q4F16GPU · WebGPUFailed—
BNB4CPU · WASMFailed—
BNB4GPU · WebGPUFailed—
Test setup & sample details
Quality gate
Finite, non-clipped, non-silent 16 kHz audio of plausible duration whose normalized 64-bin amplitude envelope has dynamic-time-warping distance at most 0.45 from the checked-in stochastic FP32/WASM reference.
Tested
2026-10-23T12:52:43.471Z
Environment
Isolated headless Google Chrome on macOS; Transformers.js 3.7.6
Runtime
Transformers.js 3.7.6
Artifact revision
1723781b8ce2d02f0400c8337be04ae8ee3d6d56
Fixture
{
  "file": "speecht5-tts.json",
  "url": "https://f005.backblazeb2.com/file/hubters-webai-public/assets/benchmarks/speecht5-tts.json",
  "size": 2177,
  "sha1": "b11a7a07ab6586b410c367c280b46c44edcff514",
  "license": "CC0-1.0",
  "reference": "64-bin absolute-amplitude envelope generated by FP32/WASM and checked into the fixture",
  "speakerEmbedding": {
    "file": "speecht5-speaker-embedding.bin",
    "url": "https://f005.backblazeb2.com/file/hubters-webai-public/assets/benchmarks/speecht5-speaker-embedding.bin",
    "size": 2048,
    "sha1": "7eca50ca7a6e271a2107f7f2db31c00d0bcdfb66",
    "values": 512,
    "sourceRepository": "Xenova/transformers.js-docs",
    "sourceRevision": "542437c29cc5273cfd29e25c77696f2b349badd7"
  }
}
Rerun when
artifact revision change; runtime version change; fixture, speaker embedding, reference, or quality-gate change; browser/backend behavior change
Per-case quality & failure details
[
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "envelopeDtwDistance": 0.221,
      "durationRatio": 0.971,
      "grade": "good"
    },
    "runtime": {
      "loadMs": 18970.3,
      "inferenceMs": 5525.5,
      "totalMs": 24635.3,
      "secondsPerAudioSecond": 1.289,
      "grade": "slow"
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": 18389.6,
      "inferenceMs": null,
      "totalMs": 21315.6,
      "grade": "failed"
    },
    "failure": "webgpu-matmul-error"
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "envelopeDtwDistance": 0.227,
      "durationRatio": 0.92,
      "grade": "good"
    },
    "runtime": {
      "loadMs": 9905.4,
      "inferenceMs": 8464.1,
      "totalMs": 18573.7,
      "secondsPerAudioSecond": 2.083,
      "grade": "very-slow"
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 8328,
      "grade": "failed"
    },
    "failure": "webgpu-session-error"
  },
  {
    "precision": "int8",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "envelopeDtwDistance": 0.349,
      "durationRatio": 0.746,
      "grade": "failed"
    },
    "runtime": {
      "loadMs": 6809.3,
      "inferenceMs": 5488.5,
      "totalMs": 12434.4,
      "secondsPerAudioSecond": 1.665,
      "grade": "very-slow"
    },
    "failure": "quality-duration"
  },
  {
    "precision": "int8",
    "device": "webgpu",
    "status": "fail",
    "quality": {
      "envelopeDtwDistance": 0.612,
      "durationRatio": 0.08,
      "grade": "failed"
    },
    "runtime": {
      "loadMs": 6002.7,
      "inferenceMs": 3796.8,
      "totalMs": 9980.4,
      "secondsPerAudioSecond": 10.786,
      "grade": "very-slow"
    },
    "failure": "quality-duration-envelope"
  },
  {
    "precision": "uint8",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "envelopeDtwDistance": 0.377,
      "durationRatio": 0.725,
      "grade": "failed"
    },
    "runtime": {
      "loadMs": 6171.6,
      "inferenceMs": 5307.8,
      "totalMs": 11617.3,
      "secondsPerAudioSecond": 1.659,
      "grade": "very-slow"
    },
    "failure": "quality-duration"
  },
  {
    "precision": "uint8",
    "device": "webgpu",
    "status": "fail",
    "quality": {
      "envelopeDtwDistance": 0.552,
      "durationRatio": 0.007,
      "grade": "failed"
    },
    "runtime": {
      "loadMs": 5783.5,
      "inferenceMs": 3068.4,
      "totalMs": 8978.8,
      "secondsPerAudioSecond": 95.887,
      "grade": "very-slow"
    },
    "failure": "quality-duration-envelope"
  },
  {
    "precision": "q8",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "envelopeDtwDistance": 0.263,
      "durationRatio": 0.957,
      "grade": "good"
    },
    "runtime": {
      "loadMs": 6389.7,
      "inferenceMs": 6237.7,
      "totalMs": 12756.9,
      "secondsPerAudioSecond": 1.477,
      "grade": "slow"
    }
  },
  {
    "precision": "q8",
    "device": "webgpu",
    "status": "fail",
    "quality": {
      "envelopeDtwDistance": 0.508,
      "durationRatio": 0.181,
      "grade": "failed"
    },
    "runtime": {
      "loadMs": 6364.6,
      "inferenceMs": 4988.2,
      "totalMs": 11504.1,
      "secondsPerAudioSecond": 6.235,
      "grade": "very-slow"
    },
    "failure": "quality-duration-envelope"
  },
  {
    "precision": "q4",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "envelopeDtwDistance": 0.276,
      "durationRatio": 0.891,
      "grade": "good"
    },
    "runtime": {
      "loadMs": 4354.5,
      "inferenceMs": 11196.6,
      "totalMs": 15744.5,
      "secondsPerAudioSecond": 2.845,
      "grade": "very-slow"
    }
  },
  {
    "precision": "q4",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": 4541.9,
      "inferenceMs": null,
      "totalMs": 7301.6,
      "grade": "failed"
    },
    "failure": "webgpu-session-error"
  },
  {
    "precision": "q4f16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "envelopeDtwDistance": 0.283,
      "durationRatio": 0.913,
      "grade": "good"
    },
    "runtime": {
      "loadMs": 4422.7,
      "inferenceMs": 11680.8,
      "totalMs": 16242.1,
      "secondsPerAudioSecond": 2.897,
      "grade": "very-slow"
    }
  },
  {
    "precision": "q4f16",
    "device": "webgpu",
    "status": "fail",
    "quality": "Not embedded",
    "runtime": {
      "loadMs": null,
      "inferenceMs": null,
      "totalMs": 4353.9,
      "grade": "failed"
    },
    "failure": "webgpu-session-error"
  },
  {
    "precision": "bnb4",
    "device": "wasm",
    "status": "fail",
    "quality": {
      "envelopeDtwDistance": 0.539,
      "durationRatio": 0.877,
      "grade": "failed"
    },
    "runtime": {
      "loadMs": 4583.4,
      "inferenceMs": 8820,
      "totalMs": 13527,
      "secondsPerAudioSecond": 2.278,
      "grade": "very-slow"
    },
    "failure": "quality-envelope"
  },
  {
    "precision": "bnb4",
    "device": "webgpu",
    "status": "fail",
    "quality": {
      "envelopeDtwDistance": 0.696,
      "durationRatio": 4.783,
      "grade": "failed"
    },
    "runtime": {
      "loadMs": 4488.6,
      "inferenceMs": 84768.6,
      "totalMs": 89398.4,
      "secondsPerAudioSecond": 4.014,
      "grade": "very-slow"
    },
    "failure": "quality-duration-envelope"
  }
]

Something not quite right?

Tell us what happened with SpeechT5. 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.