Alibaba has made the Wan3.0 artificial intelligence model for video generation available. It can create clips lasting up to 30 seconds at 1080p resolution from multiple inputs, including text, images, audio, and video, as well as PDF files, presentations, web pages, and spreadsheets.
The model had been available experimentally since the beginning of August through Alibaba Cloud’s Model Studio and Qwen Cloud platforms, before officially being opened to a broader group of users. Wan3.0 is presented as the latest member of Alibaba’s video-generation family.
What Is Changing in Practice?
Wan3.0 raises the maximum video length in a single generation process to 30 seconds, compared with 15 seconds in the previous Wan2.7 model. According to Alibaba, in longer clips the model maintains greater consistency in the appearance of characters and objects, the arrangement of elements in the scene, and animation.
The company also points to support for movements synchronized with facial expressions and multilingual audio output. Inputs are not limited to text prompts or images; content from a presentation or PDF document can be converted into video material, broadening the model’s potential use in corporate communications, marketing, and social-media content production.
Uses and Cost
Alibaba identifies marketing and corporate-communications clips, short videos, social-media content, and short dramas among the targeted uses. It also mentions potential use in training simulations for autonomous vehicles and robotic systems, without providing additional technical details about these applications in the material.
The cost varies by resolution, reaching $0.05 per second at 480p, $0.10 at 720p, and $0.20 at 1080p. As a result, producing one minute at 1080p costs approximately $12. For comparison cited in the source, Google’s Veo 3.1 model costs $0.40 per second on the standard tier.
Availability and Openness
Wan3.0 represents a shift from the approach used for previous Wan models, as Alibaba is currently keeping the model weights closed. Availability is expanding gradually through Model Studio and Qwen Cloud, with a plan to roll out an application programming interface in stages. The company also plans to offer a subscription-based consumer platform through wan.video.
Why Does This Launch Matter?
Wan3.0’s significance lies not only in doubling the video duration, but also in combining video generation from multiple enterprise sources, such as presentations and documents, with relatively longer outputs at resolutions of up to 1080p. This could matter to marketing and communications teams that work with ready-made materials and need to convert them into video, but the practical value of these capabilities will depend on the quality of the conversion and the consistency of the results, aspects for which the source provides no independent measurements.
The launch follows Alibaba’s announcement that it sold approximately $10.2 billion worth of shares in Hong Kong to fund artificial-intelligence investments. The company said it would allocate the entire proceeds from the issuance of 710 million new shares to artificial-intelligence chips and infrastructure and model development. In the latest quarter, annual net profit fell 75%, while capital expenditures rose 75% to 67.68 billion yuan. By contrast, cloud and artificial-intelligence revenue grew 45% in the June quarter to 48.44 billion yuan. Alibaba had announced at the beginning of 2025 that it intended to invest 380 billion yuan in artificial-intelligence infrastructure over three years.