Formula Ocr

OCR and recognize mathematical formulas from PDFs and images using MinerU. Converts printed or handwritten equations into structured LaTeX or text representation. Features: mathematical formula recognition from PDFs and images (.png, .jpg, .jpeg, .webp). Converts formulas to LaTeX notation. Handles complex multi-line equations, fractions, integrals, and matrices. OCR mode for scanned formula content. Use when you need to: OCR a math formula, recognize equations from images, convert formula screenshots to LaTeX, extract math from scanned documents. Use when asked: 'how do I OCR this formula', 'convert equation image to LaTeX', 'I have a photo of a math formula', 'can my agent recognize mathematical equations', 'is there a skill for formula OCR', 'turn this equation photo into text'. Powered by MinerU (OpenDataLab, Shanghai AI Lab) with advanced formula recognition capabilities. Supports a wide range of mathematical notation. Ideal for students, researchers, educators, and anyone who needs to digitize printed or handwritten mathematical content into editable LaTeX.

Install

openclaw skills install @mzlzyca/formula-ocr

Formula Ocr

Convert and extract content from .pdf / images (.png/.jpg/.jpeg/.webp) using MinerU (mineru-open-api).

Install

bash
npm install -g mineru-open-api
# or via Go (macOS/Linux):
go install github.com/opendatalab/MinerU-Ecosystem/cli/mineru-open-api@latest

Quick Start

bash
# Extract formulas from PDF (requires token)
mineru-open-api extract paper.pdf -o ./out/

# With VLM for better formula accuracy
mineru-open-api extract paper.pdf --model vlm -o ./out/

Authentication

Token required for extract and crawl:

bash
mineru-open-api auth            # Interactive token setup
export MINERU_TOKEN="your-token" # Or via environment variable

Create token at: https://mineru.net/apiManage/token

Capabilities

  • Supports local files and URLs
  • Requires token (mineru-open-api auth or MINERU_TOKEN env)
  • Supported input: .pdf / images (.png/.jpg/.jpeg/.webp)
  • Language hint with --language (default: ch, use en for English)
  • Page range with --pages (where applicable)

Notes

  • Formula recognition requires extract with token. The --formula flag is enabled by default.
  • Output goes to stdout by default; use -o <dir> to save to file
  • Binary formats (docx) require -o flag (cannot stream to stdout)
  • All progress/status messages go to stderr
  • MinerU is an open-source project by OpenDataLab (Shanghai AI Lab): https://github.com/opendatalab/MinerU