What are the best practices to make the inference FASTER? (with more gpu usage ?)

OpenTalker / SadTalker

[CVPR 2023] SadTalker：Learning Realistic 3D Motion Coefficients for Stylized Audio-Driven Single Image Talking Face Animation

https://sadtalker.github.io/

Other

11.65k stars 2.17k forks source link

What are the best practices to make the inference FASTER? (with more gpu usage ?) #735

Open AIhasArrived opened 9 months ago

AIhasArrived commented 9 months ago

I searched and found one mention of how to make things a bit faster ina comment, but I did not find the info inside a document or what not, Do you know where I can official docs about how make it faster?

Also if you have the info, can you tell me what are all the things i can do to make it faster? Please tell me everything I can do to use BIGGER GPU

johndpope commented 9 months ago

maybe this helps - https://github.com/Spycsh/xtalker

AIhasArrived commented 9 months ago

Amazing, is there a collab code version of this Xtalker? 2) How did you come to find it @johndpope ? I wondered for a long time and only now heard of this. 3) Do you know similar "faster" repos like this but for Lip Sync such as wav2lip? Thanks