Free Mathpix alternative: images and PDFs to Markdown

What it does Pix2Text ingests images or entire PDFs and reconstructs them as structured Markdown. It chains together separate models for layout analysis, table recognition, text OCR, and math formula detection and recognition to preserve reading order and formatting across 80-plus languages. The output is a single Markdown file mixing body text, LaTeX equations, and tables.

The interesting bit Instead of relying on one giant vision model, the project treats document conversion like an assembly line, delegating each content type to a specialized engine. It also recently added optional VLM backends through LiteLLM, letting you route trickier table or text tasks to closed-source models when the local pipeline struggles.

Key highlights

• Supports 80+ languages, using CnOCR for English and Simplified Chinese and EasyOCR for others
• Converts entire PDFs—not just single images—into Markdown
• Math detection and recognition models upgraded to v1.5 as of July 2025
• Optional VLM integration via LiteLLM for offloading recognition to cloud models
• Free hosted web tier available with a 10,000-character daily limit

Caveats

• The author warns that the Python toolkit is “not very user-friendly for those who are not familiar with Python”
• The free online service caps usage and explicitly discourages bulk API calls due to limited hardware
• The hosted web version currently runs newer models and outperforms the open-source local release

Verdict A solid choice if you need to batch-convert technical documents, lecture notes, or research papers into editable Markdown without a subscription. If you only convert the occasional screenshot, the free web interface is likely less hassle than orchestrating local models.