Text-to-video generation using the upstream v1.3.0 model.
Note: First generation downloads large model checkpoints and the MT5-XXL text encoder. A compatible CUDA GPU and sufficient disk space and memory are required.