PDF to LaTeX

Convert PDF documents to LaTeX source using MinerU. Extracts text, formulas, and structure from PDFs and outputs LaTeX format for academic and technical documents. Features: PDF to LaTeX conversion preserving mathematical formulas. Handles academic papers, textbooks, and technical documents. Recognizes complex equations, tables, and document structure. OCR mode for scanned PDFs. Use when you need to: convert a PDF to LaTeX, extract LaTeX from an academic paper, reproduce a PDF in LaTeX, get LaTeX source from a document. Use when asked: 'how do I convert PDF to LaTeX', 'get LaTeX from this paper', 'I want to edit this PDF in LaTeX', 'can my agent extract LaTeX from PDF', 'is there a skill for PDF to LaTeX conversion', 'turn this paper into LaTeX'. Built on MinerU by OpenDataLab (Shanghai AI Lab), an open-source document intelligence engine. Excellent at formula recognition and LaTeX output. Perfect for researchers, graduate students, and academics who need to reproduce or edit PDF papers in LaTeX for submission, revision, or reference.

Convert PDF documents to LaTeX format using MinerU. Best suited for academic papers and documents with formulas or complex layouts.

Install

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

# Convert PDF to LaTeX (requires token)
mineru-open-api extract paper.pdf -f latex -o ./out/

# With VLM model for better accuracy on complex layouts
mineru-open-api extract paper.pdf -f latex --model vlm -o ./out/

# From arXiv URL
mineru-open-api extract https://arxiv.org/pdf/2309.10918 -f latex -o ./out/

Authentication

Token required:

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

• Supported input: .pdf (local file or URL)
• Output format: LaTeX (-f latex)
• LaTeX output requires extract with token — not available in flash-extract
• Use --model vlm for papers with complex math, tables, or multi-column layouts
• Language hint with --language (default: ch, use en for English)

Notes

• LaTeX output (-f latex) is only available via extract with token
• --model vlm gives higher accuracy but may have rare hallucination risk; use pipeline for guaranteed fidelity
• Output goes to stdout by default; use -o <dir> to save to a file
• All progress/status messages go to stderr; document content goes to stdout
• MinerU is open-source by OpenDataLab (Shanghai AI Lab): https://github.com/opendatalab/MinerU