Alibaba (Qwen)
What Alibaba's Wan3.0 video model means for you
The new AI video generator makes 30-second clips with sound built in, and it's cheaper than Google's rival Veo model.
The answer
Alibaba launched Wan3.0, an AI model that makes 30-second videos with audio.
Alibaba has released a new AI video tool that can turn text, documents or even web pages into 30-second video clips complete with sound, voices and lifelike facial expressions.
What happened
Alibaba's Tongyi Lab officially launched Wan3.0 on 24 August 2026, after testing it publicly since 6 August. The launch came just days after Alibaba raised $10 billion by selling shares, in what Reuters called the largest-ever share sale of its kind by a Hong Kong-listed company, money the firm plans to put towards AI development.
Wan3.0 can generate clips up to 30 seconds long, twice the length of its predecessor, Wan2.7-Video. It offers three quality levels: 480p, 720p and 1080p (full high definition). Unlike many video generators that need a separate step to add sound, Wan3.0 creates audio, including synchronised facial movements and voices in several languages, in the same pass as the video itself. A feature called "intelligent duration" suggests how long a clip should be, and finished clips can be stretched longer afterwards.
The model also accepts a wider range of source material than before. You can feed it documents of up to 50 pages, spreadsheets, slides or web pages, and it will convert them into video. A tool called "Omni-Reference" lets you supply up to 20 reference items, such as images, video, audio, documents or web pages, and the model will keep those details consistent throughout the clip, according to TechNode.
Alibaba suggests the tool for short dramas, film, advertising, tourism and marketing content, music videos, and generating simulated footage to train self-driving cars and robots.
What it means for you
Wan3.0 is priced for high-volume use. At the API (the interface developers use to plug the model into their own apps), it costs $0.05 per second at 480p, $0.10 at 720p, and $0.20 at 1080p, working out at $12 a minute for full HD. That is half the $0.40 per second Google charges for the standard tier of its Veo 3.1 model, making Wan3.0 a notably cheaper option for anyone building video into a product.
The catch is access. Wan3.0 is a closed model, meaning you can only use it through Alibaba's API rather than downloading and running it yourself. Alibaba's earlier open-weight Wan models, which anyone can download and modify, stop at version 2.2, so this newest and most capable version isn't available in that form.
How to try it
Wan3.0 is reachable through Alibaba's Model Studio API, which is currently in preview. There is no announced consumer app for casual users; access is aimed at developers and businesses building it into their own tools.
Sources
- Alibaba launches Wan3.0 video model with 30-second generation and document input — TechNode, 24 August 2026
- Alibaba launches Wan3.0, its 30-second video model, days after raising $10bn — The Next Web, 24 August 2026