Alibaba has released Wan2.1-VACE, an open-source model designed to simplify video creation and editing. This tool is part of Alibaba’s Wan2.1 series and brings together several video processing functions in a single model.
Wan2.1-VACE allows users to work with text, images, and video to create or edit videos. It supports advanced features such as video repainting, modifying selected areas, referencing images or frames, and extending video boundaries. These tools aim to help users work more efficiently while allowing more control over video content.
With this model, users can bring still images to life by adding natural movements, replace or animate objects and people, and adjust how they appear in different scenes. The model also allows for detailed changes in specific video sections without affecting the rest of the frame.
Unified interface
Wan2.1-VACE was built with a unified interface called the Video Condition Unit (VCU), which processes different input types like text, images, and video. It also uses a Context Adapter that helps manage tasks across time and space in videos.
This model can be used in a variety of fields, including short video production, marketing, post-production for film and TV, and educational videos. By open-sourcing the model, Alibaba hopes to make advanced video tools more accessible and affordable.
Two versions of the model are now available: a 14-billion-parameter and a 1.3-billion-parameter version. Both can be downloaded for free from Hugging Face, GitHub, and Alibaba’s ModelScope platform.
Earlier this year, the company also released four other models under the Wan2.1 series and a tool for creating videos based on starting and ending frames.
Get the latest before it trends. Follow Back End News on LinkedIn, Facebook, X, YouTube, and TikTok for updates and in-depth coverage across the tech and security landscape.