an Homageworks best with VACE Canny, new workflows on the way. V0 used simple captioning, so expect increased accuracy in V1. The latest example used the new workflow.
st3ll4rV0.safetensors
st3ll4rV0.zip