This is an open-source SOTA (State-of-the-Art) video generation model. It can train a commercial-grade 11B parameter video generation model with only $200,000 (224 GPUs). Developed using the Python language and the PyTorch deep learning framework, it features fast generation speed, low resource consumption, and strong scalability. It is suitable for content creators, advertisers, and social media operators. The model weights, inference code, and the entire distributed training process have been fully open-sourced.