Wan2

Wan2 is a suite of open-source video foundation models from Alibaba Cloud delivering cinematic 720p generation on consumer-grade hardware.

Wan2 (specifically the Wan2.1 and Wan2.2 releases) represents a major shift in accessible high-fidelity video synthesis. Built on a Diffusion Transformer (DiT) architecture, the 14B parameter models handle complex motion and bilingual text rendering (English and Chinese) with professional-grade stability. The system is optimized for efficiency: the 1.3B variant runs on just 8.19 GB of VRAM, while the flagship 14B model utilizes a Mixture-of-Experts (MoE) design to maintain speed without sacrificing detail. Whether generating 5-second 720p clips from text or animating static images, Wan2 provides a production-ready pipeline under the Apache 2.0 license for developers and creators alike.

https://github.com/Wan-Video/Wan2.1