For custom dataset with a precise mask

rese1f / StableVideo

[ICCV 2023] StableVideo: Text-driven Consistency-aware Diffusion Video Editing

https://rese1f.github.io/StableVideo/

Apache License 2.0

1.39k stars 87 forks source link

For custom dataset with a precise mask #28

Open saulgooodman opened 1 month ago

saulgooodman commented 1 month ago

Hello, great work! I would like to use the stableVideo method on a custom dataset, which has a clear mask representing the foreground objects. Is it possible to skip the Neural Layered Atlas (NLA) and directly edit the foreground and background without the need for training?

rese1f commented 1 month ago

hi for now NLA is necessary to conduct editing