search

Doubiiu / CodeTalker

[CVPR 2023] CodeTalker: Speech-Driven 3D Facial Animation with Discrete Motion Prior
MIT License
515 stars 57 forks source link

Dont know this technology yet, but Before using it I need to ask a question #69

Open AIhasArrived opened 7 months ago

AIhasArrived commented 7 months ago

Is it FAST? Can you use cuda , pytorch etc? I need a fast method, I have tried video-retalking and it's SO SLOW. I need a better thing.

Doubiiu commented 7 months ago

Hi Thanks for your question. I recently tested the demo wav audio (duration=12s) using both BIWI and FLAME templates, and it took approximately 7 seconds to generate the mesh sequences on a single A100 GPU (peak GPU RAM~5G). I think it cannot be regarded as FAST, especially for real-time applications.

Yes. It was implemented with PyTorch supporting CUDA.