theEricMa / DiffSpeaker

This is the official repository for DiffSpeaker: Speech-Driven 3D Facial Animation with Diffusion Transformer
137 stars 18 forks source link

Testing on a different template? #9

Closed therealron closed 5 months ago

therealron commented 5 months ago

Hi, wonderful work! Thanks for sharing. I am trying to run inference on the model trained on Vocaset, but on a different template and am getting wrong results. All I did was change the template that was being used. my template is of the same topology as flame so I was expecting to see something reasonable at least even though its not in the training set. Can you share how I can fix this? Thanks! :)

image
therealron commented 5 months ago
image

This is what my template looks like.

therealron commented 5 months ago

Sorry I was able to solve it. I was using trimesh to read the mesh which seems to reorder the vertices for some reason. If i use igl to read the mesh I'm getting the right results. image

theEricMa commented 5 months ago

Thanks for the info!