guoyww / AnimateDiff

Official implementation of AnimateDiff.
https://animatediff.github.io
Apache License 2.0
10.6k stars 872 forks source link

Number of frames in training and inference #357

Open JunHanStudy opened 5 months ago

JunHanStudy commented 5 months ago

Thank you for your codebase! It is great. I believe the number of frames in a clip for the training and inference should be the same, say a clip of 16? What if we want to generate a video of length 32? Could we generate a clip of 16 frames and then the following clip of 16 frames? If so, how could we condition on the first clip to generate the second clip?