LightOnOCR 2
Coming soonRead a document-page image and produce editable text, which may include Markdown or mathematical notation. Long pages can be cut short by the output-token limit, and recognition errors need review.
- Released
- Jan 19, 2026
One input, different models. Compare outputs and measured speed side by side on your own device.
Model comparisons are coming with the upgraded LocalAI Tools. Explore the models below, or join the waitlist to hear when it’s ready.
Read a document-page image and produce editable text, which may include Markdown or mathematical notation. Long pages can be cut short by the output-token limit, and recognition errors need review.
Run the full text detection and recognition pipeline, or use a server detector or cropped-line recognizer separately. Inputs and verified backends vary by variant.
Read a document-page image and generate DocTags markup describing its text and page structure, including tables and formulas. The output is structured markup, not a finished PDF or guaranteed exact transcription.
Locate printed text regions in a still image. Returns axis-aligned pixel boxes and confidence scores; it does not read their text.