Prerequisites:

1) Mochi safetensor model

2) Mochi VAE model

3) T5xxl Clip Encoder (Same one if you already have it for Flux)

Download model from:

https://huggingface.co/Kijai/Mochi_preview_comfy/tree/main

For more info about node and dependencies:

https://github.com/kijai/ComfyUI-MochiWrapper?tab=readme-ov-file

How to use the workflow:

1） Download all corresponding model files up top

2)  Choose Text2Video or Img2Video as input (refer to notes)

3)  Select GGUF Q8 or lower models, and fp8 clip encoder if you  have limited VRAM

Diffuse away!