Summary
This is a solid OCR tool that gives you four engine options depending on your needs. RapidOCR, PaddleOCR, and RapidDoc run locally without API keys, while SiliconFlow taps DeepSeek's cloud model for tougher jobs. It handles scanned PDFs and images in the usual formats, pulls out Chinese and English text, and keeps the structure intact. The fallback logic is smart: if the local engine chokes, it switches to the cloud automatically. Good for processing scanned contracts, old books, or screenshots where you can't copy-paste. Version 2.5.0 by yejinlei, MIT licensed, batteries included with PyMuPDF and Pillow.
Install
npx skills add https://github.com/yejinlei/pdf-ocr-skill --skill pdf-ocr-skill
