Can you make a video from one photo?
Yes — one photo is enough for a current AI video model, because it needs an identity to draw rather than footage to edit, and a single clear image carries the whole identity.
Why one is enough
The model is not tracking you across frames; it is generating each frame from a description of what you look like. One good photo answers that. A second photo of the same person, taken in the same light, adds almost nothing.
This is also why photo quality matters more than photo count. Every frame in the output inherits whatever the single input got wrong.
What a single photo cannot tell it
Anything out of frame. The back of your head, the soles of your shoes, what your jacket does when you turn — the model invents all of it, consistently within a clip but differently between clips.
Anything obscured. Hands in pockets means the model guesses at your hands from the moment the routine takes them out, and hands are already the hardest thing it draws.
The practical rule
If the routine turns the subject around, expect the far side to be plausible rather than accurate. Pick a template that stays roughly front-on when accuracy matters — a message to a friend can absorb an invented shoulder; a video for a client cannot.