1 paper
Kesen Zhao, Jiaxin Shi, Beier Zhu +5
Real-time motion-controllable video generation remains challenging due to the inherent latency of bidirectional diffusion models and the lack of effective autoregressive (AR) appro…