According to the Daily Kotaku, theAlibabaformal open source (computing)Tongyi Wanxiang Wan2.1-VACE, a model that supports video generation and editing. A single model can simultaneously support a full range of basic generation and editing capabilities, such as text-generated video, image-referenced video generation, video redrawing, localized video editing, video background extension, and video duration extension.
It is reported that the open source 1.3B and 14B versions, of which the 1.3B version can run on consumer graphics cards.
Tongyi Wanxiang is an AliCloud Tongyi series of AI painting creation grand models that can assist humans in picture creation, and was officially launched on July 7, 2023, according to 1AI. At the September 19, 2024 Ali Yunqi Conference, Ali released the Tongyi Wanxiang video generation model. The model can generate movie-quality high-definition video and is more capable of understanding Chinese.
It is worth noting that in early January this year, Tongyi Wanxiang video generation model announced the upgrade to version 2.1, the overall performance of the model has been greatly improved, especially in the processing of complex movements, restore the real physical laws, improve the texture of the film and optimize the command to follow the significant progress. 25 February, Alibaba announced the full open source of its video generation model Wanxiang 2.1 model.