# Equation image to LaTeX

> Recognize an equation with a configured local formula model.

[Open tool](https://www.toolcabana.com/ce/math-ocr) · [PDF and documents](https://www.toolcabana.com/ce/category/pdf-and-documents)

Tool ID: math-ocr. Requested language: ce. Description language: en. Complete guide translation: no; untranslated sections use English.

## Overview (en)

Equation image to LaTeX recognizes a single cropped formula in an image and returns its LaTeX code, downloadable as formula.tex. Recognition uses the RapidLaTeXOCR model in a browser worker, which downloads about 179 MB of model files on first run. Images can be up to 8 megapixels and the output up to 512 tokens.

## Supported tasks (en)

Recognizes one cropped formula in a browser worker with RapidLaTeXOCR. Downloads about 179 MB of ONNX model assets on first run. Review every symbol and command.

## Steps (en)

1. Choose a source file. If the tool accepts multiple files, arrange them in the order you want.
2. Check the source format before running the operation.
3. Run equation image to latex, review the output, then use the available copy or download controls.

## Settings

No additional settings.

## Limitations (en)

One formula crop, up to 8 megapixels and 512 generated tokens. First model download: about 179 MB. Recognition can fail on handwriting, unusual fonts and long expressions.

## Example input

```text
A useful tool makes everyday work easier. Review the results before sharing.
```

## Privacy and connections (en)

Source files and text stay in this browser. The first run downloads public models from Hugging Face and local runtime assets from the site. Models may remain in the browser cache. No source upload to Vercel, Supabase or a conversion service is required.

## Questions (en)

### Can I convert a whole page of equations to LaTeX at once?

No. The tool expects one cropped expression per image. Crop each formula separately; long expressions that exceed 512 generated tokens stop with an error asking for a smaller crop.

### Does equation to LaTeX work on handwritten formulas?

Recognition can fail on handwriting, unusual fonts and long expressions. Results are most reliable on clean, typeset formulas, and you should review every symbol and command in the LaTeX before using it.

### Why is the first equation conversion slow?

The first run downloads about 179 MB of ONNX model files for RapidLaTeXOCR. Recognition then runs locally in a browser worker, and the time depends on your device and connection.

### Where is my input processed?

Source files and text stay in this browser. The first run downloads public models from Hugging Face and local runtime assets from the site. Models may remain in the browser cache. No source upload to Vercel, Supabase or a conversion service is required.

### What are the input limits?

One formula crop, up to 8 megapixels and 512 generated tokens. First model download: about 179 MB. Recognition can fail on handwriting, unusual fonts and long expressions.

## Related tools

- [PDF to Markdown](https://www.toolcabana.com/ce/pdf-to-markdown)
- [Document to structured JSON](https://www.toolcabana.com/ce/document-to-json)
- [Table image to CSV](https://www.toolcabana.com/ce/table-image-to-csv)
- [Invoice field extractor](https://www.toolcabana.com/ce/invoice-extractor)