Photo to video: from one still and from several

A search for “photo to video” usually means not “animate this frame” but “make me a video out of this”. The difference matters: one still gives one clip of a few seconds, and anything longer is assembled from several. This chapter covers how each of the three routes works and why putting the finished video together costs nothing.

Route one: one photo, one clip

The basic case. Your photo becomes the first frame, the model builds the rest, and the aspect ratio comes from the file itself. The length is 4 to 15 seconds in whole seconds, and the sound is created together with the picture.

The limit to understand here: one photo will not give you a ninety-second video. Not because you cannot order more, but because after five or six seconds the model starts inventing ways to fill the frame, and a second movement nobody asked for appears in the clip. A long video from photos is made differently: not as one long clip, but as several short ones.

Route two: two photos, a transition between them

A capability almost nobody knows about, because the one-button “animate” tools never have it: you can set not just the first frame but the last one too. The model invents the motion so that it lands exactly on your second photo.

What this gives in practice:

• Before and after: an empty room and a furnished one, a building site and the finished house, a face before and after a haircut.
• Two angles of the same object: instead of an orbit the model invents on its own, you set both end points.
• Two frames from the same series: what used to be “pick one shot out of twenty” becomes motion between them.

The route’s main strength is that you set both ends, so the only thing the model can spoil is the middle.

Route three: several photos, one assembled video

This is where people stumble most often. A video from several photos is not made with one request but shot by shot, and then put together into one file.

And the second thing worth asking for out loud:

How to keep the same character in every shot

The most common way a series goes wrong: three clips generated from three descriptions, with a different person in each. That is not a glitch but a direct property of generating from text: it has no character consistency, and two clips from one description are two different people.

What works:

• Start from a frame, not from text. One photo of the character is animated first, and the same photo is then used as the reference for every later shot.
• Repeat the character’s description word for word. Do not retell it in your own words in every shot: copy it.
• Frame each last frame so the next shot can start from it. Then there is no jump at the cut.

What it costs

You pay for seconds of generated video. So a video made of five four-second shots costs exactly what twenty seconds of generation costs, with nothing on top for putting them together into one file.

Attach a photo, or two at once if you want to set a transition.

What to read next

The general mechanics and the difficulty ladder: animate a photo with AI. The dose of motion within one shot: animating a photo. If a voice has to be heard in the video: a talking photo. And generating video from scratch, without a source photo, is covered in the neighbouring chapter text to video.

In the Twelver chat you can turn photos into one assembled video, one shot at a time, with sound, and join them at the end at no extra charge. Your first clip is earnable without paying: sign up, complete five first tasks and come back on the second and third day.