Source: ogwata/ndlocr-lite-web-ai: NDLOCRLite Web: ブラウザ完結型日本語OCRツール（ONNX Web Runtime使用）
Japanese OCR developed by researchers at the National Museum of Japan. It features proofreading capabilities powered by generative AI, runs on the web, and is open source.
Note that the original engine, NDLOCR-Lite, is a free OCR developed by the National Diet Library of Japan and is known for its specialized performance in recognizing printed text.
runs on the web
The description says it runs in a browser, and it looks like this code does OCR in a browser, but it relies on sending your data to third-party services like Claude to do proofreading. It’s based on a different version that runs entirely in a browser: NDLOCR-Lite Web.
Thank you. I have corrected the post.
