Blendshape Generation - All Blendshapes look the same

I’ve looked on the forums here, and I found very similar issues, but none of the solutions resolved my issue, or I just don’t know how to implement the solution.


  1. Create USDC file in Omniverse embedded blender.
  2. Import the USDC file into audio2face.
  3. Set up meshes
  4. Mesh fitting
  5. Post wrap
  6. Preview in stage
  7. All blend shapes look like my default mesh shape
  8. Export my file
  9. Go back to blender, transfer shape data, it succeeded, but all blend shapes are the exact same.
  10. Go back to audio2face, check the console, a bunch warnings I don’t really understand.


  • Graphics Cards: NVIDIA GeForce RTX 4060
  • Driver version: (up-to-date)
  • GeForce Game Ready Driver: 537.42 (up-to-date)

Generate the BlendShapes before applying the A2F pipeline. I don’t know how to even interact with the A2F pipeline, but I followed these tutorials exactly:


  • Mesh has no eyes.
  • Mesh has tongue and lower teeth, but for testing simplicity, I did not specify them in Lower Denture because I’m trying to remove any additional complexity until I get workflow operational.
  • Unlike the tutorial, when I setup character, I get a gen_openMouth mesh that is not Mark. Don’t know how to have same template mesh as video.
  • Log file:
    audio2face_blendGenLog.txt (28.3 KB)

Attached is an image showing my issue.

Solution: I had separated meshes because of the lower denture and trying to make this simple, when I actually made it much more confusing. TLDR, you CANNOT have separated meshes or it wont build right. Noted.

1 Like

I won’t claim consistent or reliable results, but I have used separate meshes for different parts of the face. I just made sure I plugged them all into the “extra meshes” (e.g. static and dynamic meshes). But the overall process frequently fails for me (e.g. a2f silently exits), so I won’t claim deep knowledge!

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.