Still lamenting the high barrier to entry for video production and the complexities of post-editing? Don't worry, the Alibaba Tongyi Wan team is back with their latest heavyweight model: the All-in-One video generation and editing model—VACE.
One of VACE's most striking capabilities is its powerful conditional video generation function. This means you only need to describe your desired scene in words, and VACE will quickly turn your imagination into reality. It's like having a dream film crew at your command, with astonishing efficiency!
VACE supports various flexible and efficient editing methods, including:
- Control object movement trajectories at will. Using the Move-Anything function, you can precisely control the movement trajectory of any object in the video. Whether you want birds to fly in graceful arcs or cars to follow a specified route, VACE can easily achieve it.
- One-click video subject replacement. The Swap-Anything function allows you to freely replace the subjects in the video. Want to turn the cheetah in a documentary into a cuddly panda? Want to instantly change the protagonist in a commercial? VACE can help you achieve this "grand illusion".
- Reference subject style transfer. Through the Reference-Anything function, VACE can learn and apply the style of your provided reference object to the video subject. Whether you want to give a live-action video an anime feel, or transform an ordinary person into the subject of an artwork, it's all possible.
- Intelligent video frame expansion. The Expand-Anything function can intelligently expand the video frame content. No more worries about limited shooting angles; VACE can help you turn "narrow" into "wide," providing a broader perspective.
- Bring static images to life. The Animate-Anything function can animate static images. Whether it's old photos or artwork, they can be revitalized under VACE's magic.
Revitalizing Old Films: VACE's Video Restoration Skills
It's worth mentioning that VACE also demonstrates impressive capabilities in video re-rendering. It supports various modes, including content preservation, structure preservation, subject preservation, pose preservation, and motion preservation. This means that even blurry, detail-lost old videos can be restored to their former glory with VACE, as if time were reversed.
The Alibaba Tongyi Wan team's VACE model, with its powerful video generation and editing capabilities, will undoubtedly greatly enhance the efficiency and fun of video creation. It's like an omnipotent video magician, capable of fulfilling users' wildest creative ideas.
Although the model is not yet officially launched, it will be open-sourced. The arrival of VACE will bring about a profound transformation in the field of video content creation.
Paper Link: https://arxiv.org/pdf/2503.07598
Open Source Code: https://ali-vilab.github.io/VACE-Page/
