How to make your pet dance in a video
You make a pet dance by giving an AI one clear photo of them standing, shot from their eye level rather than from above, and picking a full-body dance template — the model puts them on two legs and moves them through the routine.
Why pets come out better than people
Nobody has a precise mental model of how a beagle's hips work upright on two legs, so the model has enormous latitude before anything looks wrong. A human face has no such latitude — you know exactly what your friend's face does, and a frame that is two percent off registers instantly.
The result is that the same model, on the same template, produces a more convincing dancing dog than a dancing person. This is why the categories are full of animals.
The photo to take
Standing, side-on or front-on, whole animal in frame including the paws and the tail. Sitting and lying photos are the most common failure: the model has to invent the back legs before it can move them.
Get the phone down to their level. A photo taken from standing looks down on them, foreshortens the legs, and hides most of the body the routine needs.
Plain floor, plain wall, no other pets in shot. A second animal at the edge of the frame will get pulled into the generation.
The two photos that never work
A pet curled up asleep. There is no pose to extend into a standing one, and the model produces something with the wrong number of joints.
A close-up of the face. The dance templates are full-body; a head shot has nothing for them to move. A close-up is only useful in the face-swap templates, which put a face into existing footage instead.
Which template to start with
Start in the easy, forgiving category rather than the long technical one. A short simple routine tells you whether the photo works, and it costs a fraction of what a long premium render does to find out the same thing.