AI photo animation
Point AI at a still photograph and it can add motion that was never recorded — an eye that blinks, a head that turns, hair that stirs. This page is about what that involves, and what it does not.
300 free credits land in new accounts at sign-in — enough for 1 video. No card, no setup.
What the AI is actually doing
Given one photograph, the model estimates how the person in it would plausibly move, renders those few seconds, and returns them as video. It is generating frames that never existed — but doing so under the constraint of the image you gave it, which is why the result looks like your photograph rather than a drawing of someone similar.
This is also why the output is short. Sustaining invented motion for longer than a few seconds, from a single still, is where the illusion starts to fail; a brief clip keeps the generation anchored to the original pixels.
No prompt, no model list
The usual AI experience — write a prompt, pick a model, tune a setting — is absent here on purpose. You supply a photograph and nothing else; deciding how to animate it is the machine’s job, and the choices involved are not exposed as dials.
For a task like this that is a feature. The words people would put in a prompt (“subtle, natural, no distortion, keep the face the same”) are the defaults the system already assumes, and a wrong setting is the easiest way to ruin an otherwise good photograph.
What it is reliably good at
Single subjects, facing the camera, in focus. Portraits, passport-style shots, wedding photographs, pictures of one person or a couple — these come back with motion that reads as a person being alive for a moment.
It also does not care about the era of the picture. A phone portrait from this year and a studio print from fifty years ago animate on the same terms; what matters is the quality of the digital copy, not its age.
Where it falls short
Faces small in the frame, heavy blur and very wide group shots are where results disappoint — there is too little detail for the motion to hold together. And because the movement is generated rather than recorded, watching closely will always reveal that it is generated; the goal is motion that survives a normal glance, not motion that fools inspection.
Photographs where the face is partly hidden — a hat brim, a steep profile, a hand near the chin — animate less convincingly than full, open views. Choosing a picture where the face reads clearly is the one decision that genuinely changes the outcome.
Beyond old photographs
Nothing here is specific to old pictures, even if that is what most people bring. A recent portrait, a pet photograph, a picture from a wedding — the same request, the same few seconds of motion. If the face is clear and the file is sharp, the age of the subject is irrelevant.
Getting started needs a Google sign-in and nothing else: 300 credits arrive with the account, which covers 1 video, and no card is requested at any point in trying it.
300 free credits land in new accounts at sign-in — enough for 1 video. No card, no setup.
Animate a photo with AI