Alibaba (BABA, Financial) has unveiled its latest open-source video generation and editing model, Wan2.1-VACE (Video All-in-one Creation and Editing). This innovative model supports multi-modal inputs, including text, images, and videos, for video generation and offers comprehensive video editing features such as reference image or video frame generation, video re-drawing, partial editing, and extension of both scenes and duration. By integrating various video processing functionalities into a single model, Wan2.1-VACE simplifies the video creation process, enhancing efficiency and productivity.
Part of Alibaba's "Wan2.1" series, Wan2.1-VACE is expected to be the industry's first open-source model offering a unified solution for video generation and editing. Users can transform static images into videos, control object movements using specified motion paths, replace specific characters or objects, add animated effects to characters, control their poses, and convert vertical images into horizontal videos with added new elements.
The Wan2.1-VACE model is available in two versions, with 14 billion and 1.3 billion parameters, respectively. It has been released on Hugging Face, GitHub, and Alibaba Cloud's open-source community ModelScope for free download.
