a plugin of ComfyUI for Long Sonic
Sonic is a method about ' Shifting Focus to Global Audio Perception in Portrait Animation',
You can produce video as long as you want by text with this comfyUI compentent
In the ./ComfyUI/custom_node directory, run the following:
git clone https://github.com/benda1989/Sonic_ComfyUI.git
cd Sonic_ComfyUI
pip install -r requirements.txt
Download model same as Sonic, and file tree looks like this:
-- ComfyUI/models/sonic/
|-- audio2bucket.pth
|-- audio2token.pth
|-- unet.pth
|-- yoloface_v5m.pt
|-- whisper-tiny/
|--config.json
|--model.safetensors
|--preprocessor_config.json
|-- RIFE/
|--flownet.pkl
-- ComfyUI/models/checkpoints
├── svd_xt.safetensors or svd_xt_1_1.safetensors
example gpu usage