Skip to content

README_OLD.md

Latest commit

 

History

History
286 lines (208 loc) · 9.68 KB

README_OLD.md

File metadata and controls

286 lines (208 loc) · 9.68 KB
Raw

PDF Tools - ComfyUI Custom Node Package

A comprehensive ComfyUI custom node collection for PDF processing with AI-powered analysis.

📢 Important Notice: Package Split

The download functionality has moved to a separate package:

  • PDF Tools (this package): PDF extraction, OCR, AI vision processing
  • Download Tools (new package): gallery-dl and yt-dlp downloaders

If you need the download nodes (yt-dlp, gallery-dl), please install the download-tools package separately:

cd ComfyUI/custom_nodes/download-tools
.\install.ps1

Both packages work independently and can be installed together or separately.

🎉 Quick Start

Just Moved to New Install?

  1. Check Installation:

    cd A:\Comfy25\ComfyUI_windows_portable
powershell -ExecutionPolicy Bypass -File .\ComfyUI\custom_nodes\PDF_tools\check_install.ps1

  2. Install Dependencies:

    • The check script will offer to install everything
    • Or manually run: .\python_embeded\python.exe -m pip install -r .\ComfyUI\custom_nodes\PDF_tools\requirements.txt

  3. Start ComfyUI:

    • Run your normal startup script
    • Nodes will appear in the "Add Node" menu

📦 What's Included

PDF Extraction Nodes

  • PDF Extractor v08 - Advanced PDF image extraction with quality assessment and spread detection
  • PDF Extractor v09 - Latest version with enhanced image processing
  • PDF Extractor v05-v07 - Previous versions for compatibility
  • Simple PDF Extractor - Basic PDF extraction without advanced features

OCR & Layout Analysis Nodes

  • Surya OCR Layout Node - State-of-the-art multilingual OCR with layout detection
  • Surya Layout OCR Hybrid - Combined layout analysis and text extraction
  • Basic Surya / Basic Surya v02 - Simplified Surya OCR interfaces
  • Enhanced Layout Parser v06 - Advanced document layout understanding
  • LayoutLMv3 Node - Microsoft LayoutLMv3 for document AI

AI Vision & Detection Nodes

  • Florence2 Rectangle Detector - Microsoft Florence-2 for object detection and vision analysis
  • Florence2 Cropper Node - Crop images based on Florence-2 detections
  • Eric Rectangle Detector - Custom rectangle detection implementation

Remote Processing Nodes

  • PaddleOCR VL Remote - Remote PaddleOCR processing for Chinese/multilingual documents

Downloader Nodes (Moved)

Note: Download functionality has moved to the download-tools package. Install it separately if you need gallery-dl or yt-dlp downloaders.

🚀 Features

✅ PDF Extraction - Extract images from PDFs with quality scoring and duplicate detection
✅ AI-Powered OCR - Surya, PaddleOCR, LayoutLMv3 for accurate text extraction
✅ Layout Analysis - Understand document structure (headers, paragraphs, tables, figures)
✅ Vision Models - Florence-2 for object detection and image understanding
✅ Spread Detection - Automatically detect two-page spreads in scanned documents
✅ Quality Assessment - Evaluate image sharpness, contrast, and overall quality
✅ Batch Processing - Process multiple PDFs or images efficiently
✅ GPU Acceleration - Leverage CUDA for faster AI model inference

📚 Documentation

Detailed Guides (in Docs/)

  • gallery_dl_node_complete_guide.md - Gallery-dl usage guide
  • yt_dlp_node_complete_guide.md - Yt-dlp usage guide
  • SETUP_COMPLETE.md - Authentication configuration
  • Test scripts for all features

🔧 Requirements

Core Dependencies (Auto-installed)

  • PyMuPDF (fitz) - PDF processing
  • Pillow - Image processing
  • numpy - Array operations
  • opencv-python - Computer vision
  • transformers - AI models

Downloader Tools (Auto-installed)

  • gallery-dl - Web media downloader
  • yt-dlp - Video/audio downloader
  • browser-cookie3 - Cookie extraction

Optional

  • FFmpeg - For yt-dlp audio extraction (download separately)
  • Tesseract - For enhanced OCR (download separately)

� Configuration Files

Important: This repository uses template config files to protect your personal credentials.

First Time Setup

  1. Copy template files: configs/*.example → remove .example extension
  2. Edit with your credentials (usernames, passwords, API keys)
  3. Your personal configs are automatically excluded from Git

See CONFIG_SETUP.md for detailed instructions.

Your personal configuration files (without .example) will never be committed to GitHub! 🔒

�💡 Quick Examples

Download from Instagram

Node: Gallery-dl Downloader
URL: https://www.instagram.com/username/
Output: ./instagram-downloads
Enable: organize_files, use_download_archive

Download from YouTube

Node: Yt-dlp Downloader
URL: https://www.youtube.com/watch?v=VIDEO_ID
Format: best (or audio-only for music)
Output: ./youtube-downloads

Extract PDF Pages

Node: PDF Extractor v08
Input: Your PDF file
Output: ./pdf-output
Enable: quality_assessment, spread_detection

🔐 Authentication

Many sites (Instagram, Reddit) require authentication:

Option 1: Browser Cookies

  • Log into site in browser
  • Enable use_browser_cookies in node
  • Select browser (Firefox, Chrome, etc.)

Option 2: Export Cookies

  • Install browser extension "Get cookies.txt LOCALLY"
  • Export cookies while logged in
  • Save to configs/instagram_cookies.json
  • Set cookie_file parameter

📁 Project Structure

PDF_tools/
├── nodes/              # ComfyUI node implementations
├── florence2_scripts/  # Florence2 AI vision models
├── sam2_scripts/       # SAM2 segmentation models  
├── configs/            # Authentication & config files
├── Docs/               # Comprehensive documentation
└── [test directories]  # Test output locations

⚙️ System Requirements

  • Windows 10/11 (primary), Linux (should work)
  • Python 3.10+ (included with ComfyUI)
  • NVIDIA GPU with CUDA (recommended for AI models)
  • 8GB+ RAM (16GB+ for AI models)
  • 5-10GB storage for packages + models

🐛 Troubleshooting

"gallery-dl not found"

✅ Already installed! Use: .\python_embeded\python.exe -m gallery_dl

"Chrome cookies not accessible"

💡 Solution: Run ComfyUI as admin, or use Firefox, or export cookies

"Instagram/Reddit downloads fail"

💡 Solution: Export cookies from logged-in browser session

"CUDA out of memory"

💡 Solution: Close other GPU apps, reduce batch size, use smaller models

See INSTALLATION_GUIDE.md for detailed troubleshooting.

🎯 Best Practices

  1. Use Download Archives - Enable to avoid re-downloading
  2. Enable File Organization - Keep downloads sorted by type
  3. Export Cookies for Private Content - More reliable than browser cookies
  4. Start with Public URLs - Test setup before authenticated downloads
  5. Monitor GPU Usage - When using AI models

📝 Version Info

  • Gallery-dl: 1.30.9 ✅
  • Yt-dlp: 2025.9.5 ✅
  • PyMuPDF: 1.26.4 ✅
  • Transformers: 4.55.0 ✅
  • Torch: 2.7.1+cu128 ✅

🤝 Contributing

This is a working codebase! If you extend it:

  • Follow existing patterns
  • Add comprehensive error handling
  • Include debug output
  • Update documentation
  • Test with real data

📄 License

Copyright (c) 2025 Eric Hiss. All rights reserved.

This software is dual-licensed:

See LICENSE.md for full licensing terms.

Third-Party Dependencies

This project relies on numerous open-source libraries, each with its own license. See CREDITS.md for a comprehensive list of all dependencies and their licenses.

Important License Notes:

  • Some dependencies (gallery-dl, PyMuPDF) use GPL/AGPL licenses
  • Commercial use requires compliance with all dependency licenses
  • See CREDITS.md for detailed licensing information

👥 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines on:

  • Code style and standards
  • Testing requirements
  • Pull request process
  • Development setup

📧 Contact

🙏 Acknowledgments

Special thanks to:

  • The ComfyUI community for the amazing platform
  • Hugging Face for hosting models and providing ML tools
  • Microsoft Research for Florence-2 vision models
  • Meta AI for SAM2 (Segment Anything Model 2)
  • All open-source developers whose libraries make this project possible

See CREDITS.md for detailed acknowledgments.

🎉 Getting Started

  1. ✅ Check installation: Run check_install.ps1
  2. ✅ Read: INSTALLATION_SUCCESS.md
  3. ✅ Start ComfyUI: Nodes will appear automatically
  4. ✅ Test: Try a simple download
  5. ✅ Configure: Set up authentication if needed

Ready to go! All dependencies are installed. Read INSTALLATION_SUCCESS.md for next steps.

📚 Additional Documentation