Recently, Alibaba announced that its video generation model WanX2.1 will be fully open source and announced the latest video effects. This news has attracted widespread attention in the technology community, marking another major breakthrough in AI technology in the field of video generation.
In January 2025, Alibaba's Tongyi Wanxiang team launched the Wanx2.1 multimodal model. With its outstanding performance in the field of video generation, the model topped the VBench review list, redefined AI-driven visual creation standards. The WanX team announced last night that it will open source its latest WANX2.1 video generation model, which will undoubtedly promote the further popularization and application of AI technology.
It is reported that Wanx2.1 overcomes the long-standing text generation problem in AI video models for the first time, becoming the world's first model to support Chinese and English text special effects. Users only need to enter text instructions to generate dynamic videos and match them with diverse transitions, particles and other special effects. In addition, through self-developed efficient VAE and DiT architectures, the model realizes efficient encoding and decoding of infinitely long 1080P videos, significantly improving the spatial and temporal context modeling capabilities.
In terms of physical law simulation, Wanx2.1 can accurately restore complex scenarios such as collision, rebound, and cutting. For example, when generating videos of "Raindrops falling on the umbrella surface and splashing water" or "figure skaters spinning", the physical coordination and movement trajectory are in line with real physical laws, effectively solving the problems of physical distortion and stiff movements of traditional models. This technological breakthrough brings a more realistic and smooth visual experience to AI video generation.