search

AIGAnimation / CAMDM

(SIGGRAPH 2024) Official repository for "Taming Diffusion Probabilistic Models for Character Control"
145 stars 11 forks source link

No need for multi-step denoising during inference? #2

Closed warmshao closed 2 months ago

warmshao commented 2 months ago

Directly predict x0?

baopin132 commented 2 months ago

The loop is multi-step denoising. https://github.com/AIGAnimation/CAMDM/blob/561c08bbfd6c9ea5239905101057dfc791caa55f/Unity/Assets/Demo/CAMDM/Diffusion.cs#L93

warmshao commented 2 months ago

thanks!