On August 24, 2026, Alibaba officially launched the next-generation video generation model Wan3.0. The model achieves comprehensive upgrades across generation duration, versatile creation, broad referencing and real-world fidelity, producing up to 30 seconds of native video per request and, for the first time, supporting direct input of document formats such as DOC, XLS, PPT, PDF and MD. The company says it aims to more accurately reproduce the physical world and further lower the threshold for high-quality video creation.

Key Upgrades and Detailed Capabilities

Wan3.0 moved from public beta (which began on August 6) to commercial availability after more than ten days of iterations based on user feedback. Major improvements include:

Doubled Maximum Duration: Each request can generate up to 30 seconds of continuous video, double the 15-second limit of the previous Wan2.7. It supports intelligent duration recommendations and an extend function to expand narratives, making it better suited for single-take and multi-shot continuous storytelling in short dramas, commercials and music videos.
Universal Referencing and Versatile Creation: In addition to text, images, audio and video, it now supports documents (DOC/XLS/PPT/PDF/MD) and web URLs for the first time. The model can read structured content and generate corresponding videos, achieving a direct conversion "from static materials to dynamic finished footage." It supports up to about 20 reference elements, strengthening consistency of characters, vehicles and scenes across shots.
Real-World Fidelity: Significant improvements in micro-expressions, body linkage, spatial layout, physical motion and on-screen text rendering aim to reduce the "AI feel" and approach the look and feel of real shooting. It also supports native synchronous audio and multilingual dubbing.
Resolution and Output: Supports 480P, 720P and 1080P outputs, defaulting to 1080P.

API pricing is: 480P 0.3 CNY/sec, 720P 0.6 CNY/sec, 1080P 1.2 CNY/sec. From August 24 to September 23, Alibaba Cloud Bailian and Qianwen AI platforms offer a limited-time 30% discount. Users can try it via Alibaba Cloud Bailian, Wanxiang official site, Wanjing Yike, Qianwen AI platform, Qianwen Creation PC client, Qianwen App, Duiyou and IF STUDIO, among other channels. Several platforms (such as LibTV, Pollo.ai, Meitu and JD Lingjing) have integrated or are testing it.

Technical and Market Background

The Wan series was developed by Alibaba Tongyi Lab, positioned for high consistency, longer duration and multi-modal referencing in video generation. Wan3.0's document input capability is regarded by the company as an industry first, directly targeting marketing, corporate promotion, education and scaled short-drama production scenarios. Its pricing is significantly lower than some international competitors (for example, Google Veo 3.1 standard tier at about $0.4/sec), reflecting Alibaba's cost advantages in AI infrastructure.

This release follows Alibaba's completion of an approximately US$10.2 billion Hong Kong share placement (funds to be entirely invested in AI), demonstrating the company's continued investment in the video generation track.

Potential Impact: From Tool to Production Workflow

1. Impact on creators and the content industry — The 30-second native length and document input substantially lower the threshold from planning to final cut. Scenarios such as short dramas, TVC ads, cultural and tourism promotions and music MVs can achieve more efficient, scalable production. Improved consistency of characters and scenes makes the model closer to practical needs for film previsualization and continuous narratives.

2. Impact on industry competition — Against intense competition from models like