Alibaba's Wan3.0 generates AI videos up to 30 seconds long from text, images, and documents
decoder ·
Alibaba has released Wan3.0, its latest AI video generation model, in beta. It can create videos up to 30 seconds long using text, PDFs, web pages, and PowerPoint files as input, doubling the length of its predecessor Wan2.5. The model processes text, images, video, and audio simultaneously, with a single prompt accepting up to ten images, five videos, and five audio clips, and it offers tools to recommend optimal video length and extend existing videos. According to Alibaba, it improves consistency of characters, props, and spatial layouts to reduce visual drift and distortion in faces and interfaces. Wan3.0 is available through wan.video, Alibaba Cloud Model Studio, and the Qwen Cloud API in Standard and Prime tiers with per-second pricing. Alibaba targets applications ranging from film production and marketing videos to robotics and autonomous vehicle training simulations. The launch comes as the company increases AI spending, following a record share sale and a sharp quarterly profit decline due to higher AI investments.
decoder