Alibaba entered the AI video race with Wan3.0. After showcasing the exceptional capabilities of the Qwen models, the Chinese AI giant launched its new model on August 24. Notably, it was made available in public beta earlier this month.
According to the announcement, Wan3.0 can create videos up to 30 seconds long, which is almost double the capabilities of its predecessors, such as Wen2.7. As input, users can provide anything from images and videos to PDFs, PowerPoint presentations, or spreadsheets. The model can create a video out of anything.
Alibaba says the model is being used for films, short dramas, ads, tourism and music videos.
Additionally, the company mentioned that the AI model can keep people and objects more consistent during the video. The model supports video output of up to 1080p. This makes it more useful for creators who want to turn different types of content into video without having to start from scratch.
Google Veo is already a major name in AI video. It can create videos from text and images, with a strong focus on realistic scenes and movement. Wan3.0 is taking a slightly different route by supporting a wider mix of inputs.
Price could be another big point. Wan3.0 starts at about USD 0.05 per second for 480p video. The price rises to USD 0.10 per second for 720p and USD 0.20 per second for 1080p. This could make the model attractive to creators who want to make more videos without spending too much.
Wan3.0 is part of Alibaba’s wider push into AI. The launch comes soon after the company raised around USD 10 billion through a share sale to support its AI plans. For now, the real test will be quality. If Wan3.0 can deliver good-looking videos at a lower cost, Alibaba could pose a serious new rival to Google and other AI video makers.