search

wuhaozhe / audio2face_mm2023

37 stars 3 forks source link

"mesh sequence" when performing inference #2

Open Zrduan724 opened 9 months ago

Zrduan724 commented 9 months ago

Hi, thanks for your great work.

I would like to know how to prepare the "mesh sequence" input data when performing inference with unseen audio data？

Zrduan724 commented 9 months ago

Hi, thanks for your great work.

I would like to know how to prepare the "mesh sequence" input data when performing inference with unseen audio data？

More specifically, if I have one wild audio clip without mesh sequence, would the learned facial animation style only be consistent with the subject styles in the training set? Because I can only use the mesh sequence in training set.

wuhaozhe commented 7 months ago

Hi, thanks for your great work. I would like to know how to prepare the "mesh sequence" input data when performing inference with unseen audio data？

More specifically, if I have one wild audio clip without mesh sequence, would the learned facial animation style only be consistent with the subject styles in the training set? Because I can only use the mesh sequence in training set.

yes, usually during inference, we leverage mesh sequence in training set as style reference