search

showlab / Tune-A-Video

[ICCV 2023] Tune-A-Video: One-Shot Tuning of Image Diffusion Models for Text-to-Video Generation
https://tuneavideo.github.io
Apache License 2.0
4.15k stars 377 forks source link

Bad results generated from stable-diffusion-v2-1 #85

Open wanghao14 opened 10 months ago

wanghao14 commented 10 months ago

Hi, thanks for opening source this excellent work! I had tried stable-diffusion-v1-4, and the edited results were satisfactory. However, when I switched to stable-diffusion-v2-1, there was no improvement on the edited results and in fact, the results looks like noise. Have you encountered a similar situation before?

result from v1-4: sample-500 result from v2-1: sample-500

My experimental environment: cuda 11.7, Pytorch 2.0.1, ,xformers 0.0.21, accelerate 0.22.0.

ChenHoy commented 6 months ago

I have the same problem, any update?