Alibaba has rolled out the beta version of Wan3.0, an advanced video generation model that supports 30-second video length and multimodal reference inputs. This tool represents a significant leap in generative AI, integrating high-fidelity video creation, precise visual consistency, and comprehensive multimodal inputs into a single model built for professional workflows.
As Wan3.0 begins public beta testing, users can apply for testing on Alibaba Cloud’s AI development platform Model Studio and the AI-native cloud platform Qwen Cloud.
Double the video length, more control
Mainstream AI video generators typically produce clips lasting only a few seconds to 15 seconds — which was the maximum clip duration of Wan2.7-Video, the preceding Wan video model. Wan3.0’s native support for sequences up to 30 seconds per clip lets creators execute complex camera movements and continuous, unbroken shots. The model also adds an intelligent duration feature that recommends the optimal video length based on your prompts, plus video extension tools to expand narrative timelines.
Feed it almost anything
A key differentiator for Wan3.0 is its broad multimodal input. The model can process text, image, video, and audio inputs at the same time, while also handling web pages and documents like PDFs and PowerPoint presentations. That effectively lets users turn static, text-heavy data directly into dynamic video.
To tackle the visual drifting and distortion common in AI-generated media, Wan3.0 folds in high-precision visual continuity — rendering realistic faces with synchronized micro-expressions, producing natural multilingual voice outputs, and accurately presenting stable user interfaces and motion graphics.
The model is designed for a wide range of industries, from filmmaking and short dramas to social media content and business marketing. It can even help tech developers generate realistic simulation videos to train self-driving cars and robotics systems.
Wan3.0 builds on Alibaba’s Wan series of visual generation models, first introduced back in July 2023, which has gone through continuous upgrades to make image and video creation more realistic and easier to control.
