search

yerfor / GeneFacePlusPlus

GeneFace++: Generalized and Stable Real-Time 3D Talking Face Generation; Official Code
MIT License
1.59k stars 229 forks source link

Movements highly emphasized on the generated video #166

Open enrico310786 opened 7 months ago

enrico310786 commented 7 months ago

Very great work, well done!

I try to fine tune the model on my own video. The video is squared and centered on the subject's face. The generated video is of well quality but the person's movements, especially those of the neck, are highly emphasized. The person swings his neck a lot to the right and left. Is there a parameter to set on the fine tuning or in the inference step in such a way the movements are reduced?

Thanks in advance.

yerfor commented 7 months ago

Hi, you can try to set --drv_pose=static in the infer script.