Alibaba Unveils Wan3.0 with 30-Second AI Video Generation and Multimodal Inputs

Alibaba’s Wan3.0 enters public beta with 30-second video generation, multimodal inputs, improved visual consistency and tools for professional content creation.

Tabung Haji

Alibaba has launched the beta version of Wan3.0, its latest video generation model designed to produce longer, more consistent AI-generated videos from a broader range of inputs.

The new model supports video outputs of up to 30 seconds per clip, doubling the maximum duration offered by its predecessor, Wan2.7-Video.

The extended length enables creators to produce continuous shots, complex camera movements and longer narratives without relying as heavily on multiple short clips.

Wan3.0 also features an intelligent duration function that recommends suitable video lengths based on prompts, alongside video extension capabilities.

A key upgrade is its multimodal reference capability. Wan3.0 can process text, images, videos and audio simultaneously, while also accepting webpages and documents such as PDFs and PowerPoint presentations. This allows users to transform static, information-heavy materials into dynamic video content.

Alibaba said the model is designed to address common issues in AI-generated media, including visual drift and distortion. It can maintain consistency in characters, products, spatial layouts, software interfaces, motion graphics and audio while generating realistic facial expressions, micro-expressions and multilingual voices.

The model targets a broad range of applications, including filmmaking, short-form dramas, social media content, marketing and educational videos. It can also generate realistic simulations for training autonomous driving and robotics systems.

Wan3.0 is currently available for public beta testing through Alibaba Cloud’s Model Studio and Qwen Cloud, allowing users to apply for access and test its capabilities.

#businessnews