How to animate a photo with AI

What the model actually does

An image to video model treats your photo as the starting point and generates the frames that follow. The scene, lighting and appearance are already set by the picture, so the prompt does not need to describe what is visible. Its job is to explain what changes over time: who moves, what the camera does and what happens in the background.

Some models also accept a last frame. The model then builds a transition between two images, so you control both how the clip starts and how it ends.

Choosing the right photo

A model cannot add detail that is missing from the source, and blur or noise tends to get worse once things start moving. Use a sharp image where the main subject is large and not cut off by the frame.

• The face or product fills a good part of the frame
• There is room around the subject if it will move or the camera will pull back
• Even light with no blown highlights or deep shadows on the face
• The photo matches the target aspect ratio, for example vertical for Shorts
• It is a photograph, not a reproduction of a painting. With artwork the model may pan across the canvas instead of animating the subject

Picking a model

Models differ in what they do best, not only in overall quality. Use the table as a starting point for the models in the Solistra catalog. The exact duration and resolution of each version are shown on its model card before you run it.

Describing motion and camera

A reliable animation prompt answers four questions: who moves, how exactly, what the camera does and what happens around them. Name the subject. "Puts on glasses" works worse than "the woman in the photo slowly puts on sunglasses with her right hand" because the model does not have to guess who acts.

Give the camera its own sentence and one move: push in, pull out, pan, orbit, rise. If the feel matters, add pace and scale, such as "a slow push in to a close up of the face". The examples below are good templates.

• Portrait: "The girl in the photo turns her head to the camera and smiles, a light breeze moves her hair. The camera slowly pushes in to a close up. Soft daylight"
• Family photo: "The grandfather lifts his cup and takes a sip, the grandmother laughs and looks at him. Static camera. Warm window light"
• Product: "A perfume bottle stands on a stone, the camera smoothly orbits half a circle around it. Drops fall into the water nearby. The bottle keeps its shape and color"
• Landscape: "Clouds drift over the lake, light ripples cross the water. The camera slowly rises like a drone"

When to add a last frame

Use start and end frames when the ending matters: the product has to land in a hand, the hero has to face the camera, day has to turn into evening. Pick two images of the same scene with similar light and angle. The more they differ, the more likely the model is to insert a cut instead of a smooth transition.

A handy trick is to create the second frame from the first in a photo editor, changing only the detail you need. That gives the model a believable path from start to finish.

Common mistakes and fixes

Change one thing per attempt: the model, the motion description or the camera. That way you know what actually improved the result.

• Too many actions. A short clip fits one or two events, and five actions come out half done or jerky
• A prompt that fights the photo. Asking for something the scene cannot contain leads to sudden cuts. Describe motion that is natural for the image
• Hard physics. Bouncing balls and thrown objects are still difficult for video models, so simplify the action
• Warped faces and hands. Use a photo with a larger face, reduce the range of motion and avoid aggressive camera moves
• Exact counts. Models are poor at keeping numbers of objects, so "ten balloons" can easily become seven

How to do it in Solistra

Credits are reserved when the generation starts and charged after the result. If the model fails, the reserve is returned.

• Open Video and upload a photo, or send any image from your generations to video
• Describe the motion or choose a camera preset: push in, pull out, orbit, tracking shot, drone rise, handheld or dolly zoom
• Choose the model, duration and format, and check the credit price on the button
• Run it. The server shows a duration estimate before you start, and the finished video lands in your library