AnimateDIFF + ELLA

[shauray8]

Should be pretty straightforward,
AnimateDIFF adds some motion layers to UNet
ELLA works on time step aware semantic connector and prompt encoder,

In theory should work fine, let's see.