Picsart-AI-Research / Text2Video-Zero

[ICCV 2023 Oral] Text-to-Image Diffusion Models are Zero-Shot Video Generators
https://text2video-zero.github.io/
Other
3.91k stars 336 forks source link

Temporal inconsistency #44

Open rakesh-reddy95 opened 1 year ago

rakesh-reddy95 commented 1 year ago

Hi, Thanks for open-sourcing the code. I'm currently using video pix2pix and have seen that the temporal inconsistency between the frames generated. This is much worse in >16 frame videos. May I know how to fix this.

https://user-images.githubusercontent.com/20672248/236388812-5294039c-cd70-4980-9491-2ed27e8fcd2f.mp4