How an AI first dance video is made
The site sends your photo to the Vidu Q3 video model with a fixed instruction: keep both people's faces, hair and clothes exactly as they are, then have the couple slow dance closely, swaying and turning gently, in soft romantic light with a slow camera orbit. Your own words come after it, so they only need to say where and how.
With one photo of the couple, that photo is the first frame and the dance grows out of it. With two separate photos, the model treats them as references and builds a new shot with both people in it.
The result is an AI recreation, not real footage. Small details such as hands, rings or the pattern of a dress can change while the couple moves.









