Hugging Face's logo

ByteDance
/
AnimateDiff-Lightning

Text-to-Video
Diffusers
stable-diffusion
animatediff
Model card Files
xet
 Community
41

Not reproducible examples

#8
by olegchomp - opened
Discussion
olegchomp

Tried to reproduce result with "Smiling girl" in HF Space and in ComfyUI with animatediff_lightning_workflow. Image comes absolutely different, with a lot of noise.

PeterL1n
 ByteDance org

1 step will have a lot of noise, others don't have noise.
olegchomp

Tried with 4 and 8 steps.
Also example: https://youtu.be/ktKteQ8X3B0?si=qcrnu5gCEzueVbgI&t=157
PeterL1n
 ByteDance org

Use the exact setting as our comfyui example. Use Euler schedule with sgm_uniform. You can use it with ddp etc

Sign up or log in to comment