local_ai_ocr

Why we included this project

Everything DeepSeek-OCR processes here runs on the machine you install the app on, which is the whole point if you handle Vietnamese, Chinese, Japanese, or English documents too sensitive to send elsewhere. The desktop interface takes images and PDFs, lets you pick a page range for longer files, and runs a queue for batch work. Output keeps tables and layout where it can, extracts charts from inside documents as images, and exports a formatted result straight to Word. After the initial model download it makes no network calls, so your files stay on your own hardware. CPU-only machines work but run slowly, so a decent Nvidia GPU makes the difference in practice, and the project is upfront about double-checking output on important documents before relying on it.