Tutorial: Creating a Digital Human MV with Suno + Sonic + Wan2.1

I attempted to recreate an MV in ComfyUI using open-source models as much as possible. While the results from Yue and DiffRhythm didn’t quite capture the right feel, I ultimately chose Suno for the music. However, the images and videos were entirely generated using the open-source Flux, Sonic, and Wan2.1, making this workflow about 80% open-source.

Overall, I’m quite satisfied with the outcome, though image clarity is something that could be improved. I’ll continue refining the workflow and may replace different components with more advanced models and tools in the future.

The final MV has been published as a standalone video:

Workflow：https://drive.google.com/drive/folders/1zn8n46l5rYilPtyOMWQAqKmn3VhBCeE4?usp=sharing

LoRA Models

Vintage Aesthetic: https://www.shakker.ai/modelinfo/ab56a15ff374405ea78931158b3ddcfe
InContext Consistency: https://github.com/ali-vilab/In-Context-LoRA

Plugins
Image Grid Split (4-grid slicing): https://github.com/spacepxl/ComfyUI-RAVE
Sonic (Lip-sync animation): https://github.com/smthemex/ComfyUI_Sonic
Wan TeaCache & EnhanceAVideo: https://github.com/kijai/ComfyUI-KJNodes

Видео Tutorial: Creating a Digital Human MV with Suno + Sonic + Wan2.1 канала 黎黎原上咩