Wan Video

Wan Video generates video from text and from still images, with automated scene generation and motion effects, positioned for social content. Image-to-video is the more useful half in practice: you start from an asset you already control — a product shot, a frame you have art-directed — and add motion to it, rather than gambling everything on a prompt.

We track it; it is not in daily rotation here. We list it as the option for social-format clips where a source image already exists and the job is simply to make it move. If self-hosting is a requirement for you, the Wan line is worth investigating on that basis specifically — but check what is actually published, and under which licence, before you design a pipeline that assumes you can run it yourself.

Kling and Seedance are the direct comparisons and the models we would run the same brief through. Runway is the pick when you want a full editing surface rather than raw generation. Luma and Pika round out the set worth testing before committing to any of them. Wan is right when image-to-video is the core need and you want an option outside the hosted-only vendors; it is the wrong call when you need consistent characters or products across many shots, which this whole tier still handles unreliably.