Alibaba launches Wan3.0, 30-second AI video with built-in audio

What happened: Alibaba's Tongyi Lab officially released Wan3.0 on 24 August, after a public beta that ran from 6 August.

The numbers: Clips run up to 30 seconds, double the 15-second limit of Wan2.7-Video, in 480p, 720p or 1080p. API pricing is $0.05 per second at 480p, $0.10 at 720p and $0.20 at 1080p — $12 a minute. Google's Veo 3.1 standard tier charges $0.40 per second.

The details: Audio generates in the same pass by default, including synchronised micro-expressions and voices in several languages. An "intelligent duration" feature suggests clip length, and videos can be extended. New "Omni-Reference" input takes up to 20 references — images, video, audio, documents or web pages — and keeps them consistent. Documents up to 50 pages, spreadsheets, slides and web pages can all be converted straight into video.

The context: The launch came days after Alibaba's $10 billion share placement, which TechNode reported funds its AI spending. Reuters called it the largest-ever primary follow-on offering by a Hong Kong-listed company.

The catch: Wan3.0 is closed and API-only, via a Model Studio API still in preview. Alibaba's open-weight Wan line stops at Wan 2.2.

Who's affected: Alibaba points to short dramas and film, advertising, tourism, marketing and music videos as target uses, plus simulated video for training self-driving cars and robotics systems.

The rivals: Wan3.0's 1080p pricing undercuts Google's Veo 3.1 standard tier by half, positioning Alibaba as the cheaper option among leading video generators, according to The Next Web.

What's next: Wider access depends on the Model Studio API moving out of preview.