Alibaba releases open source video generation model WAN 2.1

In a move set to revolutionize the world of video generation, Alibaba has just released Wan 2.1, an open-source video generation model, marking a significant advancement in AI technology. The model offers several features, including text-to-video, image-to-video, and video editing capabilities. Notably, Wan 2.1 surpasses OpenAI's Sora in performance tests and boasts a spatio-temporal VAE architecture that allows for 2.5 times faster video reconstruction than its competitors.   One of the most striking aspects of this release is its accessibility. A consumer-friendly version allows users with standard RTX 4090 GPUs to generate five-second videos in just four minutes. This democratizes high-quality video generation, making it feasible for a broader audience, from independent filmmakers to small businesses. The models are available globally on Alibaba Cloud's ModelScope and HuggingFace platforms for academic, research and commercial use.   Alibaba emphasizes Wan 2.1's superior performance, particularly in motion smoothness and subject consistency, attributing this to its advanced architecture and extensive training on a dataset of 1.5 billion videos and 10 billion images. The model suite includes three primary variants:   Wan2.1-I2V-14B: excels in image-to-video synthesis with 480P and 720P resolutions. Wan2.1-T2V-14B: specializes in text-to-video generation, supporting both Chinese and English. This is the only model capable of producing videos in both languages.   Wan2.1-T2V-1.3B: optimized for consumer-grade GPUs, enabling rapid video generation.   This release is part of Alibaba's broader strategy to invest $52 billion in AI and cloud computing, positioning the company as a leader in these fields. The open-source nature of Wan 2.1 fosters a collaborative environment for innovation in AI video technology.   Open-Source License Alibaba has announced that WanX 2.1 will be fully open-sourced in the second quarter of 2025, including the model, training datasets, and lightweight toolkits . This move is expected to further democratize access to advanced AI video generation technology and accelerate innovation in the field. Alibaba also recently previewed its reasoning model, QwQ-Max, with plans to make it open source in the future .