Alibaba has opened beta access to Wan3.0, its latest video generation model. The system can create AI videos up to 30 seconds long and is designed to work from a wider range of source material than a simple text prompt.
What Wan3.0 Can Generate
The main shift is video length. Wan3.0 can produce clips up to 30 seconds long, which doubles the length supported by its predecessor, Wan2.5.
The model can also recommend the best video length based on the user's prompt. For existing clips, Alibaba includes an extension tool that can make videos longer.
That matters because short AI video tools often force creators to think in fragments. A longer output gives users more room for a complete scene, a product explanation, a training sequence, or a social media clip that does not end as quickly.
Inputs Go Beyond Text Prompts
Wan3.0 is built to process multiple media types at the same time. It can handle text, images, video, and audio together in one prompt.
A single prompt can include up to ten images, five videos, and five audio clips. The model can also use web pages and documents such as PDFs and PowerPoint presentations as input sources.
That means static material can become the basis for dynamic video. A document, a presentation, or a web page can provide the source content, while images, video, or audio can guide the look, movement, or reference details.
- Text can describe the intended video.
- Images can provide visual reference material.
- Video can be included as part of the prompt.
- Audio can also be supplied as input.
- PDFs, web pages, and PowerPoint files can turn static information into video content.
Alibaba Is Targeting Consistency Problems
AI-generated video can struggle with visual drift and distortion. The source material specifically points to issues that often appear in faces and user interfaces.
According to Alibaba, Wan3.0 is intended to keep reference details more consistent. Those details include characters, props, and spatial layouts.
This is an important practical goal. If a model changes a character between shots, bends an interface, or loses track of where objects are placed, the video becomes harder to use in professional work. For marketing, training, film production, or simulation footage, consistency is not just a visual preference; it affects whether the output can be trusted for the task.
Where Wan3.0 Is Available
Wan3.0 is available through the wan.video website, Alibaba Cloud Model Studio, and via API on Qwen Cloud.
Alibaba is offering two tiers. The Standard version is currently available at a 30 percent discount, while the Prime version is positioned as the faster option.
The split between a discounted Standard tier and a faster Prime tier suggests different use cases. Some users may prioritize cost while testing prompts and workflows. Others may need quicker output for production schedules or higher-volume content creation.
Why Alibaba Is Pushing Into AI Video
Alibaba is presenting Wan3.0 as a tool for several kinds of users. The company is pitching it for film production, short dramas, social media clips, marketing videos, training videos, and realistic simulation footage.
For content creators, the promise is faster production of clips from prompts and reference files. For businesses, text and images could become marketing or training videos. For tech developers, generated simulation footage could support training for autonomous vehicles and robotics systems.
The launch also fits into Alibaba's broader AI investment push. The company recently announced the largest share sale by a Hong Kong-listed company to fund its AI effort. It also reported a 75 percent year-over-year drop in quarterly profit driven by sharply higher AI investments.
Wan3.0 therefore arrives as both a product update and a signal of strategic focus. Alibaba is expanding the inputs, length, and practical targets of its AI video generation system while spending heavily to compete in artificial intelligence.