LaTeX-OCR

Why we included this project

Math-heavy writing in LaTeX or Markdown usually means retyping formulas you've seen elsewhere, which is tedious and error-prone. LaTeX-OCR (pix2tex) skips that step: point it at a screenshot or photo of an equation and it returns compilable LaTeX, using a vision transformer encoder paired with a transformer decoder. The pip package gives you several ways in: a command-line tool that handles image files or clipboard content, a PyQt5 GUI that grabs a screen region and copies the rendered result to your clipboard, and a small API with a Docker image for embedding in your own tools. It also preprocesses images to pick a sensible resolution, which helps when the input is a real-world photo rather than a clean render. Since the output is plain LaTeX, it drops straight into document pipelines and note-taking workflows, and the included training scripts let you fine-tune it on your own formula styles if the pretrained weights don't cover what you need.