Wan 3.0 is Alibaba's most capable AI video generation model to date, and its defining feature is omni-reference input: a single request can combine text, images, videos, audio, a document file and even a webpage link. If your video depends on real materials, Wan 3.0 is built for it.
The reference budget is generous: up to 10 reference images, 5 reference videos and 5 reference audio clips, plus 1 document file or 1 webpage link. Reference images preserve faces, products and styles; reference videos transfer motion; audio references guide voice and pacing; a linked webpage lets the model pull in context from an article or product page.
Output spans 480p, 720p and 1080p at any whole-second duration from 2 to 30 seconds — one of the longest single-clip ranges available. Native synchronized audio comes standard, so a 30-second Wan 3.0 clip can include dialogue and ambience without a separate audio pass.
First and last frame control works alongside the reference system for precise shot design: lock the opening frame, attach reference materials for identity, and describe the journey in the prompt. The model stitches all of it into continuous motion.
There is also Wan 3.0 Prime, a high-speed version that takes exactly the same inputs at the same quality with significantly shorter wait times — useful when you are iterating.
Start generating in one click with the Wan 3.0 generator or the Wan 3.0 Prime generator.