Skip to content

INSTALLATION_GUIDE.md

Latest commit

 

History

History
302 lines (227 loc) · 8.78 KB

INSTALLATION_GUIDE.md

File metadata and controls

302 lines (227 loc) · 8.78 KB

PDF Tools Custom Node - Installation Guide

Overview

This is a comprehensive ComfyUI custom node package that provides powerful tools for:

  • PDF Extraction & Processing - Extract images and text from PDFs
  • Media Downloading - Download images/videos from Instagram, Reddit, Twitter, YouTube, etc.
  • AI-Powered Image Analysis - Florence2 vision models for rectangle detection
  • Layout Analysis - Detect document layouts and structures
  • Image Enhancement - Modern image enhancement for better quality

What's Included

1. Gallery-dl Downloader Node

Download media from 100+ websites including:

  • Instagram (posts, stories, reels)
  • Reddit (posts, galleries)
  • Twitter/X (images, videos)
  • Imgur, DeviantArt, Artstation, and more

Features:

  • Browser cookie authentication
  • File organization by type (images/videos/audio)
  • Download archive to avoid duplicates
  • Metadata extraction

2. Yt-dlp Downloader Node

Download videos and audio from:

  • YouTube (videos, playlists, channels)
  • TikTok, Twitch, Instagram videos
  • 1000+ video platforms

Features:

  • Format selection (quality presets)
  • Audio extraction (MP3, FLAC, etc.)
  • Subtitle download and embedding
  • Playlist support

3. PDF Extractor Nodes

Multiple versions for extracting content from PDFs:

  • Extract images with quality assessment
  • OCR text recognition
  • Spread detection (book scanning)
  • Metadata preservation
  • Multiple output formats

4. Florence2 Rectangle Detector

AI-powered image analysis using Microsoft's Florence2 model:

  • Detect rectangular regions in images
  • Caption generation
  • Object detection
  • Visual question answering

5. Layout Parser Nodes

Document layout analysis:

  • Detect text blocks, figures, tables
  • Enhanced OCR with multiple engines
  • Computer vision-based layout detection

Installation Steps

Step 1: Navigate to Your Python Environment

cd A:\Comfy25\ComfyUI_windows_portable

Step 2: Install Core Requirements

Use the embedded Python to install packages:

.\python_embeded\python.exe -m pip install -r custom_nodes\PDF_tools\requirements.txt

Step 3: Install External Tools (Optional but Recommended)

3a. Install gallery-dl (for web downloads)

.\python_embeded\python.exe -m pip install gallery-dl

3b. Install yt-dlp (for video downloads)

.\python_embeded\python.exe -m pip install yt-dlp

3c. Install FFmpeg (Required for yt-dlp audio extraction)

  1. Download FFmpeg from: https://www.gyan.dev/ffmpeg/builds/
  2. Extract to C:\ffmpeg\ (or any location)
  3. Add C:\ffmpeg\bin to your system PATH
  4. Verify: ffmpeg -version

Step 4: Verify Installation

Test that packages are installed:

# Test gallery-dl
.\python_embeded\python.exe -m gallery_dl --version

# Test yt-dlp
.\python_embeded\python.exe -m yt_dlp --version

# Test PyMuPDF
.\python_embeded\python.exe -c "import fitz; print(f'PyMuPDF {fitz.__version__}')"

# Test transformers
.\python_embeded\python.exe -c "import transformers; print(f'Transformers {transformers.__version__}')"

Step 5: Configure Authentication (Optional)

For Instagram/Reddit downloads:

  1. Export cookies from your browser:

    • Use browser extension: "Get cookies.txt LOCALLY" (Chrome/Firefox)
    • Save as configs/instagram_cookies.json (Netscape format is fine)

  2. Or use browser cookies directly:

    • Set use_browser_cookies: True in the node
    • Chrome requires admin privileges on Windows
    • Firefox works without admin

For Reddit API (if using config file):

  • See Docs/reddit_app_creation_guide.py for setup instructions
  • Note: Reddit API requires OAuth and may have rate limits

Step 6: Start ComfyUI

# Start ComfyUI normally
.\run_nvidia_gpu.bat

# Or with admin privileges (for Chrome cookie access)
# Right-click run_nvidia_gpu.bat → "Run as administrator"

Quick Test

Test Gallery-dl Node:

  1. Add "Gallery-dl Downloader" node to workflow
  2. Set URL: https://www.instagram.com/janaioannaa/ (or any public profile)
  3. Set output_dir: ./test-output
  4. Run workflow
  5. Check test-output/instagram/janaioannaa/images/ for downloaded files

Test Yt-dlp Node:

  1. Add "Yt-dlp Downloader" node to workflow
  2. Set URL: https://www.youtube.com/watch?v=dQw4w9WgXcQ (example)
  3. Set output_dir: ./yt-output
  4. Run workflow
  5. Check output directory for downloaded video

Test PDF Extractor:

  1. Add "PDF Extractor v08" node to workflow
  2. Load a PDF file
  3. Set output directory
  4. Run to extract pages as images

Common Issues & Solutions

Issue: "gallery-dl: command not found"

Solution: Install with pip: python_embeded\python.exe -m pip install gallery-dl

Issue: "yt-dlp: command not found"

Solution: Install with pip: python_embeded\python.exe -m pip install yt-dlp

Issue: "FFmpeg not found" (yt-dlp)

Solution:

  1. Download FFmpeg: https://www.gyan.dev/ffmpeg/builds/
  2. Extract and add to PATH
  3. Or place ffmpeg.exe in ComfyUI root directory

Issue: "PyMuPDF not found"

Solution: python_embeded\python.exe -m pip install PyMuPDF

Issue: Chrome cookies not accessible

Solutions:

  1. Run ComfyUI as administrator (Windows security restriction)
  2. Or use Firefox instead (doesn't require admin)
  3. Or export cookies to file and use cookie_file parameter

Issue: Instagram/Reddit downloads fail

Solutions:

  1. Export cookies from logged-in browser session
  2. Place in configs/instagram_cookies.json
  3. Set cookie_file parameter in node
  4. Make sure you're logged into the site in your browser

Issue: "CUDA out of memory" (Florence2/AI models)

Solutions:

  1. Close other GPU applications
  2. Reduce batch size in node settings
  3. Use smaller model variants
  4. Enable model offloading in ComfyUI settings

Issue: Transformers version conflicts

Solution:

.\python_embeded\python.exe -m pip install transformers>=4.35.0 --upgrade

Package Size Warnings

Some packages are large and optional:

  • Surya OCR: ~1GB models (advanced OCR)
  • SAM2: ~1-2GB models (segmentation)
  • Florence2 models: ~500MB-2GB (vision models, auto-downloaded)
  • PaddleOCR: ~500MB models (Chinese/English OCR)
  • EasyOCR: ~1GB models (multi-language OCR)

These are commented out in requirements.txt - only install if needed.

Minimal Installation

If you only want specific features:

Just Gallery-dl (web downloads):

.\python_embeded\python.exe -m pip install gallery-dl browser-cookie3 requests

Just Yt-dlp (video downloads):

.\python_embeded\python.exe -m pip install yt-dlp

Just PDF extraction:

.\python_embeded\python.exe -m pip install PyMuPDF Pillow numpy

Just Florence2 (AI vision):

.\python_embeded\python.exe -m pip install transformers safetensors accelerate timm

Next Steps

  1. Review the documentation:

    • Docs/gallery_dl_node_complete_guide.md - Gallery-dl setup
    • Docs/yt_dlp_node_complete_guide.md - Yt-dlp setup
    • Docs/SETUP_COMPLETE.md - Authentication setup

  2. Test with example workflows:

    • Start with simple single-URL downloads
    • Test authentication with your accounts
    • Try batch downloads from files

  3. Configure for your needs:

    • Set up cookie files for authenticated sites
    • Create custom config files for specific sites
    • Organize download directories

Getting Help

  • Check the Docs/ folder for detailed guides
  • Review test scripts in Docs/test_*.py for examples
  • Check ComfyUI console for debug output (nodes provide detailed status)

System Requirements

  • OS: Windows 10/11 (primary), Linux (should work)
  • GPU: NVIDIA GPU with CUDA (for AI models, optional for downloaders)
  • RAM: 8GB minimum, 16GB+ recommended for AI models
  • Storage: 5-10GB for packages + models
  • Python: 3.10+ (comes with ComfyUI portable)

What's Working

✅ Gallery-dl downloads (Instagram, Reddit, Twitter, etc.) ✅ Yt-dlp downloads (YouTube, TikTok, etc.) ✅ PDF extraction with PyMuPDF ✅ Florence2 rectangle detection ✅ Browser cookie authentication ✅ File organization by type ✅ Download archives (no duplicates) ✅ Metadata extraction ✅ Debug output and error handling

Known Limitations

⚠️ Reddit API may hang with old credentials (use browser cookies instead) ⚠️ Chrome cookies require admin privileges on Windows ⚠️ Some sites require valid login cookies ⚠️ Large models (Florence2, SAM2) require GPU memory ⚠️ Transformers versions may need updates

Support & Updates

This is a custom node package. For issues:

  1. Check the documentation in Docs/
  2. Review test scripts for working examples
  3. Check ComfyUI console for detailed error messages
  4. Ensure all requirements are installed correctly

Happy downloading and processing! 🚀