how many GPU memory needed to run inference it seems RTX4080 with 24G vram is not enough

[Mohand995]

[hjadbajkdba]

@Mohand995 Hello, by default we render every 100 frames in the final stage. If your GPU memory is limited, you can modify
motion_processer.py and change 100 to 1. This will slightly increase the generation time, but it won’t affect the output quality. With this setting, the model requires only about 4 GB of GPU memory.

[Mohand995]

thank you for your reply

[Mikerhinos]

I have a 16GB GPU and going from 100 to 19 frames fits nicely 👍