Hangzhou, China — Alibaba has introduced the Wan2.6 series, the latest evolution of its visual generation models designed to bring cinematic, professional-quality video creation within reach of creators worldwide.
The new release allows users to appear in AI-generated videos as themselves—complete with their own voice—while supporting flexible multi-shot storytelling that enables richer narratives and more dynamic scenes.
At the center of the launch is Wan2.6-R2V, a new reference-to-video generation model that lets users upload a short reference video capturing a character’s appearance and voice. With simple text prompts, creators can then generate entirely new scenes starring that same person, animal, or object, while preserving distinctive visual features and vocal identity.
Multiple subjects can also be combined in a single production, opening the door to more complex storytelling and ensemble scenes.
Powered by multimodal reference generation, Wan2.6-R2V is China’s first model of its kind to seamlessly insert real people into AI-generated videos with consistent visuals and synchronized audio. This capability is expected to significantly streamline workflows for short-form drama and content creators, reducing production time while expanding creative freedom.
Beyond reference-to-video, the Wan2.6 series delivers major upgrades across Alibaba’s broader visual generation lineup. Enhancements to its text-to-video, image-to-video, and image generation models introduce intelligent multi-shot storytelling that maintains visual continuity across scenes.
Improved audio-visual synchronization and audio-to-video generation add realism through more natural dialogue timing and richer sound effects.
The models now support video outputs of up to 15 seconds, giving creators more room to develop ideas and pacing. Combined with sharper instruction-following and higher visual fidelity, Wan2.6 enables cinematic-style content that approaches professional production standards.
Image generation capabilities have also advanced. The new models support interleaved text-image outputs with stronger logical reasoning, precise artistic style control, and high-fidelity portrait generation and image editing.
With a deeper understanding of lengthy Chinese and English prompts, Wan2.6 helps creators translate nuanced ideas and artistic intent into expressive visual content.
The Wan2.6 models are available through Model Studio on Alibaba Cloud and via Wan’s official website, with planned integration into Alibaba’s flagship Qwen App.
First unveiled earlier this year, the Wan series continues to evolve rapidly, underscoring Alibaba’s push to lead in AI-driven multimedia technologies and to empower creators with tools once reserved for large studios.
