Skip to content

Ask HN: What's the best way to extract text from information dense pdfs?

1 pointrkwz2 comments
On HN

Examples for PDFs: Pitch Decks, Annual Reports etc which have text, charts, tables etc.

Comments

Obviously all the big named online AI will do this trivially. Upload file, ask for everything you want in whatever format you want. If I were doing it: https://mistral.ai/news/mistral-ocr

To do it offline due to privacy, vision enabled LLM. Biggest Gemma you can handle, qwen2.5 vl, or Mistral small. I'd probably choose mistral.

Openwebui does pdfs built in. https://docs.openwebui.com/features/document-extraction/

TBH havent tried it myself but I bet it works.

pdf2txt

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.