Show HN: Textsnap – CPU OCR in <1 seconds/page, with VLM fallback

Category: cli-tool

Tags: ocr, cli, onnx, cpu-inference, paddleocr, vision-language-model

Score: 6.3/10 (Innovation: 5, Technical: 5, Documentation: 8, Utility: 7)

Textsnap is a CLI tool that performs OCR on images, screenshots, URLs, or webpages using a CPU-optimized PP-OCRv6 model with an optional vision-language model (PaddleOCR-VL) fallback for difficult pages. It runs fully offline, supports clipboard integration, batch processing, and structured markdown output, making it a practical alternative to cloud OCR services.

Target audience: developers, devops, data engineers

Repository: https://github.com/kouhxp/textsnap · Python · MIT · 195 stars

View on Hacker News