Do you need a prompt for the Gangnam Style AI video?

If you try it in a general AI video tool, your prompt has to pin down:

• the camera: one handheld wide-angle front selfie camera, one take with no cuts;
• the place: a lit-up night street in Seoul;
• three people, each tied to their own photo;
• a close-up on each person in turn, in the trend's order, with the timing of the original clip;
• a landscape 16:9 frame.

Miss one and it stops looking like the trend.

• Faces swap or blend. With three people, most models mix features between them, worst in the close-ups.
• The camera wanders. Words like "whip pan" and "moves in" come out different every time, so the timing rarely matches the trend.
• Retries add up. Each try is a full generation, paid for whether you keep it or not.

We wrote and tested the Gangnam Style shot so you don't have to. The camera moves and timing follow the original clip, each person's own photo guides their close-up, and everyone keeps their face, hair and clothes.

You add photos, tap once and get a free 4-second preview first: the three of you together, then the middle close-up.