pdf2md-ocr 1.0.0

Simple CLI tool to convert PDFs to Markdown using Marker AI

pdf2md-ocr

Simple CLI tool to convert PDFs to Markdown using Marker AI.

Quick Start

Recommended (no installation needed):

uvx pdf2md-ocr input.pdf -o output.md

Traditional installation:

pip install pdf2md-ocr
pdf2md-ocr input.pdf -o output.md

Usage

# Convert PDF to Markdown (output same name with .md extension)
pdf2md-ocr document.pdf

# Specify output file
pdf2md-ocr document.pdf -o result.md

# Convert specific page range (page numbering starts at 1)
pdf2md-ocr document.pdf --start-page 2 --end-page 5

# Convert from page 3 to the end
pdf2md-ocr document.pdf --start-page 3

# Convert from the beginning to page 10
pdf2md-ocr document.pdf --end-page 10

# Show cache location and size
pdf2md-ocr document.pdf --show-cache-info

# Show help
pdf2md-ocr --help

# Show version
pdf2md-ocr --version

Page Range Options

• --start-page N: Starting page number (1-based, inclusive). If omitted, starts from page 1.
• --end-page M: Ending page number (1-based, inclusive). If omitted, goes to the last page.

Both options are optional and can be combined:

• Use only --start-page to convert from a specific page to the end.
• Use only --end-page to convert from the beginning to a specific page.
• Use both to convert a specific range.

Important: Page numbering starts at 1 (not 0).

First Run

The first time you run pdf2md-ocr, it will download ~2-3GB of AI models. These models are cached locally and reused for all future conversions.

To see where models are cached:

pdf2md-ocr input.pdf --show-cache-info

This will show the cache location and size after conversion. Cache locations, typically:

• macOS: ~/Library/Caches/datalab/models/
• Linux: ~/.cache/datalab/models/
• Windows: %LOCALAPPDATA%\datalab\models\

To clear the cache: Simply delete the cache directory shown in the info above, or use make clean-cache if developing locally.

Subsequent runs will be much faster since the models are already cached.

Requirements

• Python 3.10 or higher
• ~2GB disk space for AI models (one-time download)

Development

For development, a Makefile is provided with common tasks:

# Install dependencies
make install-dev

# Run tests
make test

# Run tests with verbose output
make test-verbose

# Clean build artifacts
make clean

# Clear AI model cache (frees ~3GB disk space)
make clean-cache

# Build distribution packages
make build

# See all available commands
make help

How It Works

This tool is a minimal wrapper around the excellent marker-pdf library, which uses AI models to:

1. Detect text, tables, and equations in PDFs
2. Extract content with proper formatting
3. Convert to clean Markdown

License

GPL-3.0-or-later

This project is licensed under the GNU General Public License v3.0 or later to comply with the marker-pdf library license (GPL-3.0-or-later).