Alibaba's Wan3.0: AI videos up to 30 seconds, from text to PowerPoint

The launch of Wan3.0 comes as Alibaba ramps up its AI spending significantly. The company just announced the largest share sale by a Hong Kong-listed company to fund its AI push, and last week reported a 75 percent year-over-year drop in quarterly profit driven by sharply higher AI investments. This new model is a key part of that strategy, targeting a wide range of commercial applications.

Alibaba is positioning Wan3.0 for professional and business use cases. These range from speeding up film production and generating short dramas to creating marketing and training videos from text and images. The company is also courting tech developers who could use it to produce realistic simulation footage for training autonomous vehicles and robotics systems, areas where consistent visual details are crucial.

The model's focus on maintaining consistency in characters, props, and spatial layouts addresses a known weakness in AI-generated video, which often suffers from visual drift and distortion. By improving on this and supporting more complex inputs like multi-page documents, Alibaba is attempting to make its model a more practical tool for converting static business data into dynamic video content.