Repository navigation
Inbuilt Image Parser with Local OCR (Unlimited OCR, PaddleOCR etc.) #146
Description
Feature Request
- add local ocr option on cpu only (using unlimited ocr, or paddle ocr) engines (optional installation)
- add image parser so calling it "anydoc" actually makes more sense
- example: anydoc [.jpg/.png/.webp/url]
Activity
cedhuf commentedon Sep 1, 2026
Same here, that would be great. I can't use it in my app for now because of this.
calicoagent commentedon Sep 28, 2026
For anyone blocked on this in the meantime: the Python binding raises
NeedsOcrErrorwith the page numbers that need OCR, so you can OCR only those pages locally and keep the text pages going through anydoc. I wired that up with Tesseract (English, Japanese and Thai by default, CPU only, nothing hosted) in bon5co/anydoc-serve. Disclosure: it is mine. There is a Tesseract vs RapidOCR comparison indocs/ocr-benchmark.mdif you are choosing an engine. It is not a built-in parser, but it gets you unblocked today.