Abstract: Transformer-based video generation models have demonstrated significant potential in content creation. However, the current state-of-the-art model employing “ 3 D full attention” encounters ...