1 paper
Wenshuo Peng, Gongxuan Wang, Tianmeng Yang +4
Recent text-to-video generation models have made remarkable progress in visual realism, motion fidelity, and text-video alignment, yet they still struggle to produce socially coher…