notes-2-LaTeX

Convert images of mathematical notes and equations into clean LaTeX with a Streamlit interface powered by OpenAI vision.

Live app: https://notes-2-latex.streamlit.app/

Features

• Upload .png, .jpg, or .jpeg images.
• Extract only mathematical formulas as LaTeX.
• Convert a complete photographed page into a compilable .tex document.
• Preview and copy generated LaTeX from the Streamlit UI.
• Download full-document output for use in Overleaf or another LaTeX editor.
• Uses gpt-4o-mini for image understanding.

Included apps

• Foto2LaTeX (dual: FÓRMULAS o DOCUMENTO completo).py — recommended version with formula-only and full-document modes.
• Foto2LaTeX (solo FÓRMULAS → LaTeX).py — lightweight formula-only version.

Requirements

• Python 3.10+
• Streamlit
• OpenAI Python SDK
• Pillow
• An OpenAI API key

Install dependencies:

pip install -r requirements.txt

Configuration

Set your API key as an environment variable:

export OPENAI_API_KEY="your-key"

On Windows PowerShell:

$env:OPENAI_API_KEY="your-key"

For Streamlit Cloud, store the key in the app secrets as OPENAI_API_KEY.

Never commit API keys to the repository.

Run locally

Clone the project:

git clone https://github.com/MarioIbago/notes-2-LaTeX.git
cd notes-2-LaTeX
pip install -r requirements.txt

Run the dual-mode app:

streamlit run "Foto2LaTeX (dual: FÓRMULAS o DOCUMENTO completo).py"

Or run the formula-only version:

streamlit run "Foto2LaTeX (solo FÓRMULAS → LaTeX).py"

How it works

1. The uploaded image is normalized and encoded for the vision request.
2. The selected mode asks the model either for raw equation LaTeX or for a complete compilable document.
3. The response is returned to the Streamlit interface for preview, copy, or download.

License

See LICENSE.