How to turn a photo into a video
There are three ways to turn a photo into a video: a camera move over the still, a text prompt describing what should happen, or a reference clip whose movement is transferred onto the subject — and only the third gives you a specific dance.
One: move a camera over the still
The oldest method and the cheapest. Pan and zoom across the photo, maybe with a parallax layer or two. Nothing in the scene moves; the camera does.
Use it for a slideshow. It cannot make anybody dance, because it never generates a frame the photo did not already contain.
Two: describe what should happen
Give a model the photo and a sentence — "she turns and waves at the camera" — and it generates a clip that fits both. This is what most AI video apps mean by image-to-video.
It is enormously flexible and completely unrepeatable. The same prompt twice gives two different clips, and "do the dance from that video" is not something a sentence can specify.
Three: transfer movement from a reference clip
Hand the model a clip of the movement alongside the photo. It reads the pose out of the clip and applies it to your subject. This is motion control, and it is what a dance template is.
The trade is control for freedom: you get exactly the routine in the reference, every time, and you cannot get anything that is not in the library unless you write a prompt instead.