Subscribe
Sign in
PDF2MD — Convert your PDFs to markdown With AI OCR
Trieve PDF2MD is an intelligent OCR tool that makes PDF content accessible and ready for AI.
Using cutting-edge models like GPT-4o-mini and Gemini-flash-1.5, we convert PDFs into clean, structured Markdown with remarkable precision and cost-efficiency.
6
Replies
David Dennison
Wow was pretty fast and accurate from the small go that I gave it go a minute ago! And that is a much better price point!
Share
People would think OCR should be a solved problem by now, but it's so difficult to do it right. This is hands down the best I've seen so far... and the example on constellation projects is just mind-blowing. Also love the focus on PDFs to Markdown.
Maker
Hi!
I'm Nick, lead engineer/CEO of Trieve. As search and RAG infrastructure providers, we were tired of seeing our customers paying $$$$ for OCR when working with documents in our LLM pipeline. The reality is that every PDF needs to become markdown before it hits your vector DB or RAG system.
After trying a bunch of solutions, we found they were either:
- Expensive (looking at you, big cloud providers 👀)
- Inaccurate with complex layouts
- Too slow/unathletic for production use
We saw OmniAI's Tyler Maran's Zerox and were blown away. There are a lot of players and money in chunking. Seeing vision models like GPT-4o-mini cranking out enterprise-grade OCR at 1/10th the cost was wild. We dove in.
We've already processed millions of pages for our own products, so we've decided to make it available as an API. Super simple to integrate - just POST your PDF and get back clean markdown. You can also give it a spin at pdf2md.trieve.ai.
Would love feedback from fellow developers! Happy to jump on a call to help with integration or answer any questions: pdf2md@trieve.ai
Cheers,
Nick K
P.S. It's in Rust
@skeptrune
https://pdf2md.morethan.io - pdf2md already exists in Rust and in js - what is this demo adding besides putting the Trieve promotion on it? I have integrated this advanced pdf to structured layout in https://airesearch.js.org/functi... -- it costs 0$ bc it runs in users front end and does not requrie an expensive Rust server or OCR.
