This is the FancyVideo vae_3d_16_frames model (https://huggingface.co/qihoo360/FancyVideo), but with its keys pruned and modified to be readily loadable by ComfyUI (or other AnimateDiff-compatible backends).
There is a FancyVideo key added to make the model easily distinguishable for either AnimateDiff-like models. It's value can be ignored, but it is a tensor of size(0) with value of 1.0 to distinguish it as version 1.0.
