Binghuan Wu

Streaming audio-driven character animation can finally handle abrupt tempo changes and user edits in real-time, thanks to a diffusion-forcing approach that prioritizes both responsiveness and long-term coherence.