VEnhancer ComfyUI Extension

<div align="center">

ComfyUI extension for VEnhancer: A powerful video enhancement model that supports spatial super-resolution, temporal interpolation, and AI-guided refinement.

Features • Installation • Quick Start • Documentation •

</div>

Features

• 🎥 High-Quality Video Enhancement

  • Spatial super-resolution (up to 8x upscaling)
  • Temporal super-resolution through frame interpolation
  • AI-guided video refinement with text prompts

• 🚀 Flexible Processing Options

  • Single GPU inference for standard workloads
  • Multi-GPU support for large-scale processing
  • Adjustable enhancement parameters
  • Custom text prompting

• 🛠️ ComfyUI Integration

  • Intuitive node-based workflow
  • Real-time preview support
  • Progress tracking
  • Batch processing capabilities

Installation

Prerequisites

• ComfyUI installed and running
• Python 3.10 or higher
• CUDA-capable GPU with at least 12GB VRAM (24GB+ recommended)

Setup

1. Install in ComfyUI custom nodes directory:

cd ComfyUI/custom_nodes/
git clone https://github.com/vikramxD/VEnhancer-ComfyUI-Wrapper
cd VEnhancer-ComfyUI-Wrapper

2. Install dependencies:

uv pip install setuptools
uv pip install -e . --no-build-isolation

Documentation

Available Models

| Model | Description | Download | |-------|-------------|----------| | v1 (paper) | Creative enhancement with strong refinement | Download | | v2 | Better texture preservation and identity consistency | Download |

Core Parameters

Enhancement Settings

{
    "up_scale": 4.0,      # Spatial upscaling (1.0-8.0)
    "target_fps": 24,     # Target frame rate (8-60)
    "noise_aug": 200,     # Refinement strength (50-300)
    "solver_mode": "fast" # "fast" (15 steps) or "normal"
}

Model Configuration

{
    "version": "v2",      # Model version (v1/v2)
    "guide_scale": 7.5,   # Text guidance strength
    "s_cond": 8.0,       # Conditioning strength
    "steps": 15          # Inference steps (fast mode)
}

Troubleshooting

Common issues and solutions:

1. CUDA Out of Memory

   • Reduce up_scale value
   • Use multi-GPU processing
   • Process in smaller chunks

2. Slow Processing

   • Enable solver_mode="fast"
   • Use multi-GPU setup
   • Reduce video resolution

Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Based on VEnhancer by Jingwen He et al. If you use this extension in your research, please cite:

@article{he2024venhancer,
  title={VEnhancer: Generative Space-Time Enhancement for Video Generation},
  author={He, Jingwen and Xue, Tianfan and Liu, Dongyang and Lin, Xinqi and 
          Gao, Peng and Lin, Dahua and Qiao, Yu and Ouyang, Wanli and Liu, Ziwei},
  journal={arXiv preprint arXiv:2407.07667},
  year={2024}
}

---

<div align="center"> Made by VikramxD </div>