Added models: - Wan2.2 diffusion models: T2V, Animate 14B, T2V Low Noise - Wan2.2 LoRAs: Lightning 4-step, Lightx2v, Relight, Distill - Text encoders: UMT5-XXL (fp8 and bf16 variants) - CLIP Vision H for Wan workflows - InsightFace buffalo_l model for face detection - IP-Adapter Plus SDXL VIT-H - RealESRGAN x4 upscaler - Wan 2.1 VAE Updated file permissions to executable for all Python files and documentation. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
ComfyUI Model Installer + PMA Utils
A comprehensive ComfyUI extension pack featuring model management, video processing utilities, and custom nodes.
Features
Model Management
- One-click model installation from the ComfyUI interface
- Real-time download progress in a dedicated window
- Automatically organizes models into their correct folders
- Resumes interrupted downloads
- Skips already downloaded models
Video Processing Nodes
- Eye Stabilizer: Fixes eye glitching in video motion transfer workflows (Wan2.2, etc.)
- Character Swap: ControlNet-based character face swapping
- Temporal smoothing and enhancement tools
System Utilities
- Automatic shutdown monitoring with queue detection
- Activity tracking to prevent premature shutdowns
- ComfyUI session management tools
Installation
- Navigate to your ComfyUI custom_nodes directory:
cd ComfyUI/custom_nodes
- Clone or copy this extension:
git clone https://github.com/yourusername/ComfyUI-ModelInstaller.git
Or manually copy the ComfyUI-ModelInstaller folder to your custom_nodes directory.
- Restart ComfyUI
Usage
- Launch ComfyUI
- Click on Model Manager → Install Models in the top menu bar
- A dialog window will open
- Click Start Download to begin downloading all models
- Monitor the progress in real-time
- Close the window when complete
Requirements
Core Requirements
- wget must be installed on your system
- Linux/Mac: Usually pre-installed
- Windows: Install via chocolatey:
choco install wget
For Eye Stabilizer Node
- MediaPipe (for face detection):
pip install mediapipe==0.10.21- If not installed, node falls back to basic mode without face tracking
Models Included
The extension downloads the following models:
- ControlNet Union SDXL - Advanced control model
- GFPGAN Face Restore - Face restoration model
- CLIP Vision SDXL - CLIP vision encoder
- GonzalomoXL Checkpoint - Main checkpoint model
- IP-Adapter FaceID Plus v2 - Face identity preservation
- SwinIR 4x Upscaler - Image upscaling model
- EpiC Negative Embedding - Negative prompt embedding
- Deep Negative Embedding - Additional negative embedding
Adding More Models
To add more models to the download list, edit model_downloader.py:
MODELS = [
{
"url": "https://example.com/model.safetensors",
"path": "folder/filename.safetensors",
"name": "Display Name"
},
# Add more models here
]
Troubleshooting
wget not found error:
- Install wget on your system
- Linux:
sudo apt-get install wgetorsudo yum install wget - Mac:
brew install wget - Windows:
choco install wget
Downloads fail:
- Check your internet connection
- Verify the URLs are still valid
- Ensure you have sufficient disk space
Custom Nodes Documentation
Eye Stabilizer
Fixes eye glitching, jittering, and unnatural blinking in video motion transfer workflows.
Full Documentation: EYE_STABILIZER_README.md
Setup Guide: SETUP_EYE_STABILIZER.md
Technical Details: EYE_STABILIZER_IMPLEMENTATION.md
Features:
- Dense eye landmark detection (478 facial points via MediaPipe)
- Temporal smoothing using Kalman filtering
- Blink detection and synthesis
- Eye region enhancement
- Debug visualization
Quick Start:
Video → RMBG → [Eye Stabilizer] → DWPose/Depth → WanVaceToVideo
Test Installation:
python test_eye_stabilizer.py
Character Swap
ControlNet-based character face swapping for pose/scene transfer.
Status: Basic implementation (preprocessors only)
Planned: Full IP-Adapter FaceID integration
Project Structure
ComfyUI-ModelInstaller/
├── __init__.py # Main entry point
├── model_downloader.py # Model download system
├── shutdown_monitor.py # Auto-shutdown utilities
├── character_swap_node.py # Character swap node
├── eye_stabilizer_node.py # Eye stabilizer node (NEW)
├── test_eye_stabilizer.py # Test suite (NEW)
├── README.md # This file
├── EYE_STABILIZER_README.md # Eye stabilizer docs
├── SETUP_EYE_STABILIZER.md # Setup guide
├── EYE_STABILIZER_IMPLEMENTATION.md # Technical docs
└── js/
└── model_installer.js # Frontend UI
License
MIT