Alibaba's Wan 3.0 brings several video-generation input modes under one model identifier on Vercel's AI Gateway. It can work from text, images, reference media, or specified opening and closing frames. The release notes describe clips up to 30 seconds at 30 frames per second, with audio in the output by default.
A single identifier does not make every input combination valid. Opening-and-closing-frame conditioning cannot be combined with reference inputs, and audio references require hosted URLs. Applications can consolidate model routingSending each request to a model chosen by the difficulty of the task, rather than using one model for everything.Full definition → while still checking the constraints of the chosen generation mode.
Terms in this piece · Glossary
model routing — Sending each request to a model chosen by the difficulty of the task, rather than using one model for everything.
