Alibaba has stepped up its innovation game with the release of Wan2.1-VACE. This all-in-one open-source model promises to transform the landscape of video creation and editing. Developed as part of Alibaba’s Wan2.1 AI series, this cutting-edge model is now publicly available, aiming to make powerful video tools more accessible to creators across industries.
A Unified Solution for Video Creation and Editing
Wan2.1-VACE (Video All-in-one Creation and Editing) introduces a revolutionary approach to video production. Unlike traditional models that rely on multiple tools, this model integrates diverse functionalities into a single system. Users can now generate and edit videos using text, images, or video as inputs, all through one interface.
The model enhances flexibility and creative control by offering tasks like video repainting, object modification, spatiotemporal extension, and selective frame editing. It allows for the easy transformation of static visuals into dynamic stories.
Advanced Features That Bring Creativity to Life
One of Wan2.1-VACE’s standout features is its ability to animate static images by adding natural movements. It also allows the inclusion or removal of elements within specific areas of a video without disturbing the rest of the scene. Users can even control motion, adjust poses, and recolour subjects with advanced editing tools.
Another impressive capability is horizontal expansion. Creators can stretch vertical images into horizontal video formats, auto-filling the new space with relevant content using AI referencing. Additionally, characters or objects can be replaced with animated, pose-controlled alternatives driven by sample references.
Breakthrough Technologies for Smarter Video Workflows
At the core of Wan2.1-VACE lies the Video Condition Unit (VCU), a unified processing interface for multimodal inputs. Alibaba has implemented a Context Adapter system that understands time and space dimensions in video, allowing seamless task switching and flexible editing.
The model’s adaptability makes it suitable for various industries, from social media and marketing to filmmaking and education. Whether it’s for post-production visual effects or rapid content creation, Wan2.1-VACE promises efficiency and scalability.
Open-Source Access for Everyone
Recognizing the high computational cost of training video AI models, Alibaba is releasing Wan2.1-VACE in two versions: a powerful 14B-parameter and a lighter 1.3B-parameter model. Both versions are now available for free download on Hugging Face, GitHub, and ModelScope via Alibaba Cloud’s open-source platform.
This move aligns with Alibaba’s broader vision of democratising AI tools. Following its February release of four open models and last month’s launch of a frame-based video generation model, the company has now surpassed 3.3 million downloads across platforms.
As one of the first global tech leaders to open-source large-scale video models, Alibaba is positioning itself at the forefront of next-gen visual innovation, fostering a new era where powerful AI is within everyone’s reach.
