Paper Markdown Extractor

A small Qt desktop GUI for converting academic PDFs into clean Markdown. It extracts:

• Native PDF text with heading heuristics
• Tables, using PyMuPDF table detection
• Images into an adjacent images/ directory
• Display equations cropped as broad row images for higher visual fidelity, including right-edge equation numbers, with text math blocks available as a fallback
• Optional OCR fallback for scanned pages when Tesseract is installed
• PyMuPDF4LLM layout engine for difficult multi-column reading order
• Integrated Markdown preview with local images, equation crops, and tables rendered from the generated output

The fast path is built for digitally generated academic papers. A 10-page text PDF should usually finish well under one minute on a modern laptop. Scanned PDFs are slower because OCR requires page rendering.

Setup

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt

Optional OCR engine for scanned PDFs:

brew install tesseract

Run

python -m paper_md_extractor

Choose a PDF, choose an output folder, then click Convert. Enable PyMuPDF4LLM layout when a paper has difficult multi-column reading order or complex document structure. The app previews the generated Markdown and writes:

• <paper-name>.md
• images/ with extracted page images

Build the macOS app

The macOS package is built with pyside6-deploy. Its configuration in pysidedeploy.spec bundles PyMuPDF, including its native extensions and MuPDF libraries, and uses Nuitka 2.8.10 because the older 2.7.11 release cannot compile NumPy with Python 3.13.

From the project root, create or update the environment with all application dependencies. pyside6-deploy invokes python -m pip internally, so install pip into uv-created environments before running the build:

uv sync
uv pip install --python .venv/bin/python pip
.venv/bin/python -c "import pymupdf, pymupdf4llm; print(pymupdf.__version__)"

Build the app:

.venv/bin/pyside6-deploy -c pysidedeploy.spec -f

The resulting application is deployment/quick_text.app. Test it locally:

open deployment/quick_text.app

To create a ZIP suitable for distribution:

mkdir -p dist
ditto -c -k --sequesterRsrc --keepParent deployment/quick_text.app dist/Quick-Text-macos.zip

Screenshots

Notes

PDFs do not reliably encode equations, reading order, or table structure semantically. This app favors a fast deterministic extraction pipeline and keeps the output editable. For image-only scanned papers, enable OCR and expect slower conversion.