Alibaba Cloud has officially launched Wan3.0, the latest version of its AI video-generation model, as the Chinese technology giant expands its push into generative AI. The model can generate videos of up to 30 seconds in a single run and, for the first time in the Wan series, can accept documents such as DOC, XLS, PPT, PDF and Markdown files as inputs. Alibaba says the upgrade is designed to improve instruction following, consistency across shots and audio quality.
Wan3.0 had been available in public beta since early August, with Alibaba continuing to refine the model based on feedback from creators and enterprise users. The commercial launch comes days after Alibaba raised roughly $10 billion through a Hong Kong share placement, with the company directing the proceeds toward its expanding artificial-intelligence investments. The timing highlights the growing importance of AI infrastructure and model development to Alibaba’s broader technology strategy.
Alibaba Expands Wan Video Generation Capabilities
Wan3.0 represents a significant expansion of Alibaba’s video-generation platform. The most visible upgrade is the ability to produce clips lasting up to 30 seconds in a single generation, compared with the 15-second limit associated with the previous Wan2.7 model. The model can generate video at resolutions of up to 1080p and is designed to maintain greater consistency throughout longer sequences.
Alibaba says the model has improved its ability to follow complex instructions while preserving character details, objects, spatial layouts and motion graphics across shots. It also supports audio generation, including multilingual voice output and more synchronized facial expressions, according to reporting on the launch.
Wan3.0 At a Glance
| Feature | Wan3.0 |
|---|---|
| Developer | Alibaba Cloud |
| Model | Wan3.0 |
| Maximum video length | 30 seconds |
| Maximum reported resolution | 1080p |
| Document inputs | DOC, XLS, PPT, PDF, Markdown |
| Other inputs | Text, images, audio, video and web pages |
| Public beta | Early August 2026 |
| Official launch | August 24, 2026 |
| Primary uses | Marketing, social content, short films, training and business video |
| Access | Alibaba Cloud Model Studio, Qwen Cloud and staged consumer access |
The ability to accept documents could make Wan3.0 particularly relevant to businesses rather than limiting it to entertainment and social-media creators.
Documents Can Now Become Video
One of Wan3.0’s most notable changes is its expanded input capability. Instead of requiring users to start with a text prompt, image or video reference, the model can work from information contained in documents.
A user could potentially provide a presentation, spreadsheet, PDF or Markdown file and have Wan3.0 turn the underlying information into a video. This creates a bridge between conventional business documents and AI-generated visual content.
For companies, the capability could simplify the production of marketing materials, product explainers, training videos and internal communications. A presentation that previously required a designer, video editor and voice-over workflow could potentially become the starting point for an automated video-production process.
DOCUMENT-TO-VIDEO WORKFLOW
DOC / XLS / PPT / PDF / MD
↓
Wan3.0 Analysis
↓
Understands Structure
↓
Visual + Audio Generation
↓
Up To 30 Seconds
↓
Marketing / Training / Social Video
The feature also reflects a broader direction in generative AI: models are increasingly expected to understand multiple forms of information rather than operating from text prompts alone.
Longer Videos Could Improve AI Storytelling
The move from 15-second to 30-second generation gives creators more room to build a coherent sequence without stitching together as many separate clips.
Longer generation can be especially useful for short advertisements, product demonstrations and narrative scenes. However, longer clips also make consistency more difficult because the model has more frames and actions to maintain.
Alibaba said it improved consistency across shots during the public-beta period. The company also highlighted improvements in instruction following, audio quality and video editing capabilities.
Wan3.0 Compared With Earlier Wan Models
| Capability | Earlier Wan Generation | Wan3.0 |
|---|---|---|
| Single-generation duration | Up to 15 seconds for Wan2.7 | Up to 30 seconds |
| Text input | Yes | Yes |
| Image references | Yes | Yes |
| Audio/video references | Supported in multimodal workflows | Supported |
| Document input | Limited/not available | DOC, XLS, PPT, PDF, MD |
| Cross-shot consistency | Improving | Further upgraded |
| Audio quality | Supported | Improved |
| Maximum reported resolution | — | Up to 1080p |
The longer generation window is therefore only one part of the upgrade. Alibaba is positioning Wan3.0 as a broader multimodal creation system capable of converting different types of source material into finished video content.
Alibaba Targets Businesses as Well as Creators
The document-input capability suggests that Alibaba is targeting professional workflows alongside traditional AI-video creation.
Marketing teams could use product documents to generate promotional videos. Companies could turn presentations into short explainers, while training departments could transform technical material into visual lessons. The same system could also support social-media content and short-form entertainment.
Alibaba’s positioning therefore extends beyond the typical text-to-video creator market.
Industry users cited during the beta period said the model was useful for maintaining product appearance, logos and materials in advertising content, while other users highlighted its potential for short-film production and large-scale content creation.
Wan3.0 Pricing Puts Pressure on Video AI Rivals
Alibaba is also competing on price. According to reporting by The Next Web, Wan3.0 is priced at $0.05 per second at 480p, $0.10 per second at 720p and $0.20 per second at 1080p. At the highest resolution, that translates to approximately $12 for one minute of generated video.
| Resolution | Reported Wan3.0 Price | Approx. Cost Per Minute |
|---|---|---|
| 480p | $0.05/second | $3 |
| 720p | $0.10/second | $6 |
| 1080p | $0.20/second | $12 |
Pricing is becoming increasingly important as AI-video generation moves from demonstrations toward commercial production. Lower costs allow companies to generate more variations, conduct more creative experiments and produce larger volumes of short-form content.
However, price alone does not determine competitiveness. Video quality, consistency, generation speed, editing controls and reliability can all affect the actual cost of producing usable content.
Alibaba’s AI Push Is Getting More Expensive
Wan3.0 arrives as Alibaba is committing significant resources to artificial intelligence. The company’s recent $10 billion Hong Kong share placement was intended to support growing AI investment.
The broader financial picture shows the scale of that commitment. Alibaba’s cloud and AI revenue increased 45% to 48.44 billion yuan in the June quarter, while AI model services reached more than 16 billion yuan in annualized recurring revenue, according to The Next Web. At the same time, quarterly net profit fell 75% as capital expenditure increased 75% to 67.68 billion yuan.
ALIBABA'S AI INVESTMENT STORY
Cloud + AI Revenue
│
└── 48.44B yuan
↑ 45%
AI Model Services
│
└── >16B yuan
Annualized recurring revenue
Capital Expenditure
│
└── 67.68B yuan
↑ 75%
Net Profit
│
└── ↓ 75%
These numbers illustrate the trade-off Alibaba is making: increasing investment in AI infrastructure and products in anticipation of larger future revenue opportunities.
Wan3.0 Marks a Shift for Alibaba’s Open-Source Strategy
There is another notable change surrounding Wan3.0. Alibaba previously gained attention for releasing several Wan models with open weights. Wan2.2, launched in July 2025, was the last Wan flagship released that way, according to The Next Web. Wan3.0 is being offered as a closed model.
That shift could signal a stronger focus on commercial monetization. Rather than distributing the complete model for developers to run independently, Alibaba can monetize access through its cloud infrastructure and APIs.
The change also comes as major AI companies increasingly balance open model distribution with the costs of operating large-scale AI services.
Access Is Still Being Rolled Out
Wan3.0 is not yet being offered as an unrestricted consumer service everywhere. Access has been staged through Alibaba Cloud Model Studio and Qwen Cloud, while wider API availability is rolling out. A consumer-oriented service at wan.video is also being positioned as a members-only platform.
Alibaba is offering a temporary 30% discount on Wan3.0 API pricing on selected platforms between August 24 and September 23.
The staged rollout allows Alibaba to manage demand while monitoring the computing resources required to generate longer, higher-resolution videos.
Competition in AI Video Generation Is Intensifying
Wan3.0 enters a market that includes models from Google and other major AI companies, as well as increasingly capable systems from Chinese AI developers.
Alibaba’s combination of longer video generation, document understanding and competitive pricing gives it a distinct commercial proposition. But independent comparisons will ultimately determine how Wan3.0 performs against competing systems.
The model had not yet been independently benchmarked at the time of the latest reports, meaning many of the performance claims surrounding consistency and quality remain based on Alibaba’s own assessments and early user feedback.
The Bigger Picture
Wan3.0 shows how AI video generation is moving beyond simple text-to-video prompts toward multimodal business workflows. The ability to turn documents, presentations and spreadsheets into videos could make generative video useful to a much broader group of companies, including marketing, training and communications teams.
For Alibaba, the launch also fits into a much larger AI investment strategy. The company is spending heavily on cloud infrastructure, AI models and commercial applications while seeking new revenue streams from those investments. Wan3.0 therefore serves not only as a creative AI product but also as another test of whether Alibaba can convert its large AI spending into recurring commercial demand.
Looking Ahead
The immediate test for Wan3.0 will be whether creators and businesses continue using it after the initial launch period and whether its longer generation window produces consistently usable results. Pricing, API availability and the ability to turn complex documents into accurate videos could determine whether the model gains traction among enterprise customers rather than remaining primarily a tool for experimentation.
Over the longer term, Alibaba’s AI video strategy could become part of a wider ecosystem connecting its cloud services, Qwen models and content-generation tools. If Wan3.0 can combine lower production costs with reliable long-form consistency and document understanding, it could help push AI-generated video from a creative novelty toward a routine business-production tool.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
