Fetching the paper…

Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization · Around