Animating a photo versus restyling it: two different jobs

A motion action returns about five seconds of video showing movement that was never in the picture. A photo style returns one new still: the same moment, the same pose, remade in a different look. They cost different amounts, come back at different speeds, and go wrong in different places. Choosing between them is a question of where the result ends up.

What each one actually produces

The distinction is easy to lose because both start with a photograph and both come back changed, so it is worth being concrete.

A motion action invents time. Your photograph is one frozen instant, and the model generates the seconds either side of it: a wave, a walk, a turn, a dance. Nothing in that movement happened. There is no footage of it hiding anywhere. The output is a short clip, about five seconds, at 720p.

A photo style invents surface. The moment stays exactly where it was, the pose does not change, nobody moves. What changes is everything around and over it: the grain and palette of a particular kind of film, the lighting and staging of a red carpet, the glossy vinyl of a figurine on a display base. The output is a single new picture.

The action library is split along this line, and every action page states which of the two it returns rather than leaving you to assume.

What each one asks of your photograph

This is where the two diverge most, and it is the part worth knowing before you choose.

Motion is demanding about the crop. If an action moves the legs, the legs need to be in the picture, because a model that never saw them has to build them from nothing and invented limbs are the fastest route to a result that looks subtly wrong. How much of the body the photo needs works through this action by action.

Motion is also demanding about the background. As a body moves, the boundary between person and background has to be redrawn in every frame. A plain wall makes that trivial. A patterned rug or a railing makes it hard, and the smearing that follows is one of the most recognisable failures in the whole category.

A restyle is far more forgiving on both counts, because nothing has to move. A tight crop is fine. A busy background is often simply replaced by the style. What a restyle is demanding about instead is the face: it is being rebuilt under a new look, and the more of it the model can read clearly, the more the result stays the person you started with. Even, front-on light is worth more here than resolution, which is true of both halves and surprises people every time. Choosing a photo that will animate well covers the shared checks.

Where each one fails

Neither is more reliable than the other in general. They are unreliable about different things, and knowing which is which stops you blaming the wrong part.

Motion fails at hands, and at distance travelled. Hands are the known weak point across every tool in this category, for reasons that have nothing to do with which company built the model. Why hands look wrong in animated photos goes into that properly. And the further a body moves through space the photograph never showed, the more room there is for the face to drift away from the person you started with. A small gesture holds a likeness almost perfectly. A full routine asks for far more invention, and some of that invention shows.

A restyle fails by flattening toward its own convention. A style is a set of expectations about how a face is lit, shaped and finished, and a heavy one can pull features toward that convention rather than keeping them. The pose is untouched and the likeness is not necessarily safe. The heavier the transformation, the more true this is, which is why a subtle film treatment tends to keep a person more recognisably themselves than a figurine does.

Both of these are properties of the job rather than faults you can diagnose out of a good photograph. The library’s notes on the limits say so without hedging.

Cost and time, briefly

A restyle costs meaningfully less credit than a clip and comes back sooner, because one image is far less work than several seconds of video where every frame has to agree with the last. That difference is real enough to shape how you experiment: running the cheaper half first tells you how the model reads your particular photograph, its face, its light, its background, before you spend at the clip rate to find the same thing out.

No figure appears on this site. What it costs to make an AI video from a photo explains the credit model and where the current numbers live, and how long it takes covers the wait.

Choosing by where the result ends up

The fastest way to decide is to ignore both technologies and ask where the thing is going.

A message, a story, a post, a group chat. A clip. Motion is what stops a thumb, and five seconds is exactly the length those surfaces want.

A printed card, a framed gift, a profile picture, anything on paper. A still. A clip cannot be printed, and a restyled picture can be sent, posted and printed alike.

Something for someone who will look at it properly. Usually a still, and usually a restrained one. A heavy style is a novelty and reads as one. A quiet film treatment of a good portrait reads as a picture.

You are not sure the photograph is any good source material at all. Start with the cheaper half and find out.

The short version

Animating invents time and asks for the right crop and a readable background. Restyling invents surface and asks for a readable face. One returns roughly five seconds of video, the other returns a picture. Hands and travelled distance are where motion drifts; the style’s own conventions are where a restyle drifts. Pick by where the result is going, and let that decide which half of the library you are shopping in.