diffusion noise modify

[ErwinKC]

Hi, it's a great job, hope you have time to answer my simple question, where can I modify the Gaussian noise parameter in the inference stage or sampling stage and if I change input to a image or video, dose the model have ability to generate a video through this image or generate a awesome video though the existing poor quality video? Thanks

[maxin-cn]

Hi, it's a great job, hope you have time to answer my simple question, where can I modify the Gaussian noise parameter in the inference stage or sampling stage and if I change input to a image or video, dose the model have ability to generate a video through this image or generate a awesome video though the existing poor quality video? Thanks

Thanks for your interest.
LatteT2V is a pure text-to-video generation model that currently does not support image-to-video, nor can enhance the quality of input video.

You can change the Gaussian noise in here.