JAM: Joint Appearance-Motion Representations for Enhanced Motion Generation in Models VideoJAM introduces a framework that improves motion coherence in video generation by learning a joint appearance-motion representation. This framework uses a modified training

→ View original post on X — @askalphaxiv