Hello! I just started playing with Audio2Face. I’m following this tutorial: BlendShape Generation in Omniverse Audio2Face - YouTube
My first attempt I did it on the allison model and everything worked great. My second attempt I have been trying to use the blender rain character model (studio[.]blender[.]org/characters/5f1ed640e9115ed35ea4b3fb/v2/). Here are the steps I took and the weird outcome I’m getting:
- I modified the rain character in blender to separate the eyes and then exported it as USD
- I imported the USD file in A2F and followed the character transfer steps as shown in the video
- When doing the Blendshape Generation, I first did “Preview in Stage” and everything looked good. I then exported the blendshape generation and was able to generate the USD and usdSkel file.
- When dragging the usdSkel file into the stage, at first it seemed like nothing happend, but then I scaled the xform by 100 and there it was. 100x scale matched the original mesh size.
- Then I tried to do the Blendshape Conversion. It succeeded, but when I played the audio it seemed like the usdSkel mesh wasn’t doing anything. Zooming into the mouth, I can actually tell that it was moving very very little.
- Exporting the weights as json and opening that json file showed that there were in fact weights in there but many were showing very small values (i.e. 7.454361821146449e-05 or smaller)
Any ideas why this is happening?