1 paper
Qijun Gan, Chenwei Zhang, Meiguang Jin +2
Real-time long-form digital-human generation relies on causal models to extend audio-visual content while preserving subject appearance and audio-video synchronization across succe…