RedbeardNZ / SeedVR2_comfyUI

ComfyUI-SeedVR2_VideoUpscaler

Official release of SeedVR2 for ComfyUI that enables Upscale Video/Images generation.

📋 Quick Access

• 🆙 Note and futur releases
• 🚀 Updates
• 🎯 Features
• 🔧 Requirements
• 📦 Installation
• 📖 Usage
• 📊 Benchmarks
• ⚠️ Limitations
• 🤝 Contributing
• 🙏 Credits
• 📄 License

🆙 Note and futur releases

• Improve FP8 integration, we are loosing some FP8 advantages during the process.
• Tile-VAE integration if it works for video, I have test to do or if some dev want help, you are welcome.
• 7B FP8 model seems to have quality issues, use 7BFP16 instead (If FP8 don't give OOM then FP16 will works) I have to review this.

🚀 Updates

2025.06.30

• 🚀 Speed Up the process and less VRAM used (see new benchmark).
• 🛠️ Fixed leak memory on 3B models.
• ❌ Can now interrupt process if needed.
• ✅ refactored the code for better sharing with the community, feel free to propose pull requests.
• 🛠️ Removed flash attention dependency

2025.06.24

• 🚀 Speed up the process until x4 (see new benchmark)

2025.06.22

• 💪 FP8 compatibility !
• 🚀 Speed Up all Process
• 🚀 less VRAM consumption (Stay high, batch_size=1 for RTX4090 max, I'm trying to fix that)
• 🛠️ Better benchmark coming soon

2025.06.20

• 🛠️ Initial push

🎯 Features

• High-quality Upscaling
• Suitable for any video length once the right settings are found
• Model Will Be Download Automatically from Models

🔧 Requirements

• A Huge VRAM capabilities is better, from my test, even the 3B version need a lot of VRAM at least 18GB.
• Last ComfyUI version with python 3.12.9 (may be works with older versions but I haven't test it)

📦 Installation

1. Clone this repository into your ComfyUI custom nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler.git

1. Install the required dependencies:

load venv and :

pip install -r ComfyUI-SeedVR2_VideoUpscaler/requirements.txt

install flash_attn/triton, 6% faster on process, not a mandatory.

pip install flash_attn
pip install triton

or

python_embeded\python.exe -m pip install -r flash_attn

check here from https://github.com/loscrossos/lib_flashattention/releases and https://github.com/woct0rdho/triton-windows

1. Models

   Will be automtically download into : models/SEEDVR2

   or can be found here ( MODELS )

📖 Usage

1. In ComfyUI, locate the SeedVR2 Video Upscaler node in the node menu.

1. ⚠️ THINGS TO KNOW !!

temporal consistency : at least a batch_size of 5 is required to activate temporal consistency. SEEDVR2 need at least 5 frames to calculate it. A higher batch_size give better performances/results but need more than 24GB VRAM.

VRAM usage : The input video resolution impacts VRAM consumption during the process. The larger the input video, the more VRAM will consume during the process. So, if you experience OOMs with a batch_size of at least 5, try reducing the input video resolution until it resolves.

Of course, the output resolution also has an impact, so if your hardware doesn't allow it, reduce the output resolution.

1. Configure the node parameters:

   • model : Select your 3B or 7B model
   • seed : a seed but it generate another seed from this one
   • new_resolution : New desired short edge in px, will keep ratio on other edge
   • batch_size : VERY IMPORTANT!, this model consume a lot of VRAM, All your VRAM, even for the 3B model, so for GPU under 24GB VRAM keep this value Low, good value is "1" without temporal consistency, "5" for temporal consistency, but higher is this value better is the result.
   • preserve_vram : for VRAM < 24GB, If true, It will unload unused models during process, longer but works, otherwise probably OOM with

📊 Benchmarks

7B models on NVIDIA H100 93GB VRAM (values in parentheses are from the previous benchmark):

3B FP8 models on NVIDIA H100 93GB VRAM (values in parentheses are from the previous benchmark):

NVIDIA RTX4090 24GB VRAM

⚠️ Limitations

• Use a lot of VRAM, it will take all!!
• Processing speed depends on GPU capabilities

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

Please make sure to update tests as appropriate.

How to contribute:

1. Fork the repository
2. Create your feature branch ( git checkout -b feature/AmazingFeature )
3. Commit your changes ( git commit -m 'Add some AmazingFeature' )
4. Push to the branch ( git push origin feature/AmazingFeature )
5. Open a Pull Request

Development Setup:

1. Clone the repository
2. Install dependencies
3. Make your changes
4. Test your changes
5. Submit a pull request

Code Style:

• Follow the existing code style
• Add comments for complex logic
• Update documentation if needed
• Ensure all tests pass

Reporting Issues:

When reporting issues, please include:

• Your system specifications
• ComfyUI version
• Python version
• Error messages
• Steps to reproduce the issue

🙏 Credits

• Original SeedVR2 implementation

📜 License

• The code in this repository is released under the MIT license as found in the LICENSE file .