Alibaba’s Wan 3.0 Is Here: The AI Video Model That Turns Documents Into 30-Second Movies
On August 24, 2026, Alibaba officially launched Wan 3.0 (also called Wan3.0 or 通义万相 3.0), its most advanced AI video generation model yet. The release arrives just one day after the Chinese tech giant announced a roughly $10–10.2 billion Hong Kong share placement—the largest primary follow-on offering by a Hong Kong-listed company—with proceeds dedicated to accelerating AI investment.
This isn’t just another incremental upgrade. Wan 3.0 doubles the maximum clip length of its predecessor, accepts an unusually wide range of inputs (including PDFs, PowerPoint decks, spreadsheets, and web pages), and is rolling out across Alibaba’s own platforms and third-party tools on the same day.
Wan 3.0 entered public beta on August 6, 2026, through Alibaba Cloud Model Studio and Qwen Cloud under the model ID wan3.0-video. Early users already applied it to short dramas, advertising, tourism promotion, and music videos.
Today’s launch expands access: the previous invitation-style gate has largely lifted, full API availability is rolling out, and the model is now live on multiple platforms including fal.ai, Envato’s AI video tools, Sousaku AI, Morphic, and Alibaba’s consumer-facing surfaces such as wan.video and Qwen Creation apps.
Native 30-second single-pass generation
Most AI video models still top out at 5–15 seconds and require stitching. Wan 3.0 generates up to 30 seconds in one continuous shot—twice the limit of Wan 2.7. An “intelligent duration” feature automatically chooses an appropriate length based on the prompt so short scenes aren’t padded. Video extension is also supported for longer narratives.
Omni-Reference: Documents, Spreadsheets, and Web Pages as Input
This is the model’s biggest differentiator. Alongside text, images, audio, and video, Wan 3.0 accepts PDFs, PowerPoint presentations, Excel files, Markdown, and live web pages. Feed it a slide deck or a research report and it can turn the static content into a coherent video. Up to around 20 reference assets can be used in a single generation for stronger character, product, style, and spatial consistency.
Reality-grade rendering and synchronized audio
Alibaba emphasizes improved visual continuity, realistic faces with micro-expressions, stable on-screen text and UI elements, and more natural real-world motion. Audio (including multilingual voice) is generated in the same pass as the visuals rather than added afterward.
Output options
Resolutions: 480p, 720p, and 1080p (no confirmed native 4K at launch). Common aspect ratios are supported.
API pricing is charged per second of generated video:
A full minute at 1080p therefore costs around $12—noticeably lower than several Western models such as Google’s Veo 3.1 standard tier. Domestic Chinese pricing is listed in RMB and varies slightly by region.
Importantly, Wan 3.0 is closed-source and API-only. Unlike the earlier Wan 2.1 and 2.2 models, no weights have been released.
Wan 3.0 arrives at a moment when Chinese AI labs have quietly built strong positions in video generation while much of the Western conversation remains focused on language models. By combining longer coherent clips with the ability to ingest business documents directly, Alibaba is targeting both creative professionals and enterprise teams that need to turn reports, decks, and data into video without heavy post-production.
The model is already being positioned for short-form drama and film production, marketing and advertising, educational content, social media, tourism promotion, and even simulation footage for robotics and autonomous driving training.
Alibaba’s aggressive AI spending—reflected both in the recent earnings impact from capital expenditure and in the $10 billion raise—signals that the company intends to stay competitive in the global generative media race. Wan 3.0 is the most visible product of that strategy so far in 2026.
Whether you’re a marketer who wants to turn a PowerPoint into a polished promo, a filmmaker experimenting with longer single takes, or a developer building video features into an application, Wan 3.0 is now available on multiple platforms and ready to test.
The era of 30-second, document-aware AI video has officially begun.
