Open Edanaw opened 1 month ago
Thank you for your interest in our work! The reason for the slowness is partly that we use a larger model (SDXL and two Controlnets) and high resolution 1080p, and partly that the code only inference one frame at a time. Since it only relates to the first frame when generating the other frames, you can speed up the inference by inference different frames on different GPUs.
Thank you for your interest in our work! The reason for the slowness is partly that we use a larger model (SDXL and two Controlnets) and high resolution 1080p, and partly that the code only inference one frame at a time. Since it only relates to the first frame when generating the other frames, you can speed up the inference by inference different frames on different GPUs.
Thanks a lot
VideoGeneration main.py generate video speed is very slow. make V100 spend 3 hours generate video