pymupdf-layout 1.28.0

PyMuPDF Layout turns PDFs into structured data 10× faster than vision-based tools using AI trained on PDF internals, not images. CPU-only. No GPU required.

PyMuPDF Layout

PyMuPDF Layout is a fast and lightweight layout analysis Python package integrated with PyMuPDF for clean, structured data output from PDF. It's fast, accurate and doesn't need GPUs like vision-based models.

While other tools train machine learning models on rendered page images, PyMuPDF Layout trains Graph Neural Networks directly on PDF internals. This gives us accuracy at 10× the speed utilizing CPU-only resources.

Features

• 📚 Structured data extraction from your documents in Markdown, JSON or TXT format
• 🧐 Advanced document page layout understanding, including semantic markup for titles, headings, headers, footers, tables, images and text styling
• 🔍 Detect and isolate header and footer patterns on each page

Usage

PyMuPDF Layout works alongside PyMuDF4LLM's to_markdown method. Once PyMuPDF Layout is activated just use to_markdown and PyMuPDF Layout will work behind the scenes to analyse documents and deliver improved results.

You can also get a JSON or TXT format of the data with to_json or to_text.

Extract Structured data

import pymupdf.layout
import pymupdf4llm
source = "your.pdf"
doc = pymupdf.open(source)
md = pymupdf4llm.to_markdown(doc)
json = pymupdf4llm.to_json(doc)
txt = pymupdf4llm.to_text(doc)

Try It!

Try PyMuPDF Layout on our PyMuPDF website.

Documentation

See the PyMuPDF Layout documentation page for more.