This is a Python wrapper for Chrome's built-in screen-ai OCR engine. This engine is extremely fast compared to other alternatives (Tesseract, etc.) and very accurate (particularly for extracting text; less so when dealing with complex layouts such as tables and forms). However, it is only available through Chrome/Chromium. The magic of this wrapper is that it allows you to call the screen-ai library directly from Python (using ctypes), without having to open browser windows.
It works on Windows (chrome_screen_ai.dll), Linux (libchromescreenai.so), and should be easy to modify it to work on MacOS.
Lastly, it supports both PDFs and images (JPG, PNG, WebP, BMP, TIFF, GIF).
Quick start
To install this library, simply clone it and then install it from the local folder:
pip install -e . # install locro download # (optional) one-time: copy library + models from Chrome locro ocr document.pdf # process a PDF locro ocr photo.jpg --text # process an image
See GUIDE.md for the full user guide, including installation, CLI, and API documentation.
See CHROME_SCREEN_AI_DLL.md for technical details on how the library interface was reverse-engineered.