Research
NVIDIA and academic team propose 3D motion and video co-generation for arbitrary rigs
The research connects a separate 3D motion model supporting arbitrary skeletons to a pretrained video model through cross-attention layers, training on nearly 200K pairs. It also recovers camera placement so motion can be replayed in 3D scenes.
Every outlet that ran it 1 article
Hugging Face PapersResearchLead
Joints, Camera, Action! Co-Generating 3D Motion and Video for Arbitrary Rigs