Open Lidong1211 opened 1 year ago
You can refer to the following repos: https://github.com/kenwaytis/faster-SadTalker-API https://github.com/Spycsh/xtalker
You can refer to the following repos: https://github.com/kenwaytis/faster-SadTalker-API https://github.com/Spycsh/xtalker Thank you for your reply. 'faster-SadTalker-API' simply speeds up the Face Enhancer .'xtalker' Optimized only the machines that use CPUs.
A lightweight face renderer might be helpful. You can refer to https://github.com/OpenTalker/SadTalker/discussions/457.
I found that the Face Renderer loop takes a lot of time to perform the first step of each execution using the GPU memory (guessing it's not enough memory and performing a memory release), and very little time for subsequent reasoning. So I tried to manually release the memory before each loop, but this release process takes a lot of time (batch_size=20, cost=1.5s), and the subsequent inference process takes a small amount of time (batch_size=20, cost=0.7s), newbie, I don't know if there is a way to optimize this piece!