🚨 THE OCR TOOL BUILT FOR THE LLM ERA JUST DROPPED

@whitewhoadie

32.6K followers · usually 18.7K views a post

Views

153.1K

Likes

6.3K

Comments

66

Shares

1.1K

SoundTikTok sound · original sound - whitewhoadie

Promoted product

olmOCR

Toolkit for linearizing PDFs for LLM datasets/training - allenai/olmocr

Why it worked

The video effectively highlights a niche problem (messy document conversion for LLMs) and presents a clear solution (olmOCR) with a direct link to access it. The use of on-screen text and a concise explanation makes the value proposition immediately understandable.

Summary

The video introduces olmOCR, an OCR tool designed for the LLM era that converts PDFs, scans, and images into structured Markdown files. It highlights the tool's ability to handle various document types and complex layouts, making it useful for research papers, legal filings, and financial reports.

Structure

1Introduction of olmOCR
2Explanation of its purpose (OCR for LLM era)
3Demonstration of its capabilities (converting various formats to Markdown)
4Highlighting its use cases (research papers, legal, financial)

Original caption

🚨 THE OCR TOOL BUILT FOR THE LLM ERA JUST DROPPED olmOCR turns messy PDFs, scans, and images into clean, structured Markdown that models can actually understand. Ships with: • Handles tables, equations, handwriting & multi-column layouts • Preserves natural reading order • Works on old scans, figures, insets, headers & footers • Fixes the “first mile” problem in RAG/document AI Research papers, legal filings, financial reports — no more broken text. 👉 https://github.com/allenai/olmocr Who’s fixing their document pipeline today? Drop a 🔥#greenscreen

More from @whitewhoadie

Product placement

The video prominently features the olmOCR tool, showcasing its interface and explaining its features. The GitHub repository link is displayed on screen and mentioned in the transcript.

Call to action

Check out the GitHub repository: https://github.com/allenai/olmocr

On-screen text

🚨 THE OCR TOOL BUILT FOR THE LLM ERA JUST DROPPED olmOCR 2 https://olmocr.allenai.org/ Features: • Convert PDF, PNG, and JPEG based documents into clean Markdown • Support for equations, tables, handwriting, and complex formatting • Automatically removes headers and footers • Convert into text with a natural reading order, even in the presence of figures, multi-column layouts, and insets • Efficient, less than $200 USD per million pages converted (Based on a 7B parameter VLM, so it requires a GPU) Benchmark olmOCR-Bench: We also ship a comprehensive benchmark suite covering 7,000 test cases across 1,400 documents to help measure performance of OCR systems. https://github.com/allenai/olmocr

Transcript

WHAT'S UP GUYS So I found an OCR it's built for the LLM era. Okay, so this means it's basically optimal character recognition. And uh, so you can scan stuff that's like handwritten PNGs, JPEGs, you turn them all into markdown files. Very fun, very fun guys. So anything research papers, legal filings, financial reports, blah blah blah blah blah. Fucking yeah.