AI Animation From One Image: A Practical Guide

Animating one image with AI means giving a video model your picture as the first frame and saying what should move. The model generates the frames that follow, so you get a short clip, 3 to 30 seconds by model. In AIGE it is From a picture in the Video editor, and the price shows before you press Make.

What animating one image means here

Here is the job. You have one picture: a photo, an illustration, a render, something you made in AIGE. A video model opens a clip on that picture and generates everything after it from your words. The model pages call the picture the first frame. Kling's own how-to says the video adds frames that were not in your photo, so you review the result for recognizability. What you get is new footage that begins at your picture.

Two things follow from that. You do not draw or key the frames yourself: you give the model one picture, or two if you want it to land somewhere, and it generates the frames. And the clip does not repeat a movement you already have on video. That job is Copy a movement, covered near the end.

There is also a plainer route that uses no video model. A picture on the editor's timeline can take a slow camera move (Push in, Pull out, Pan left or Pan right) from the clip's Motion & animation section. The editor zooms or slides the picture itself, so nothing inside it moves.

Where to do it in AIGE

In AIGE this is From a picture, in the Make tab of the Video editor on the laptop canvas, and turn a photo into a video with AI walks through every step, including the optional Change it first and End on a picture switches.

Which start pictures respond

Kling's how-to, published on 30 September 2026 for Kling's own tools, starts before any prompt: decide what should move and what must stay recognizable. Its advice on the picture, as of 28 October 2026:

• A person filling the frame. A small change of expression or posture gives the clip a focus.

• A wide scene. It suits motion in the surroundings, such as water or plants.

• A drawing, an illustration or an AI-generated image. Kling says these can work. For an illustration, name the traits to keep, such as the line work or the shading.

• A face or a product that must survive. Kling lists a changed face or object as a known failure and suggests a clearer image plus naming the features to preserve.

Pictures made in AIGE sit on your board and can be the start picture directly. File limits differ by source. Kling's how-to asks for JPG, JPEG or PNG, at least 300 pixels on each side and at most 10 MB. fal's Seedance 2.5 page lists eight formats, among them WebP and HEIC, up to 30 MB. fal's pages for Kling V3, H3 Max and Omni Flash state no limits, and we found none in AIGE's form. If an upload fails, check format and size first.

Wording the motion

Kling's guides and fal's model pages agree: the prompt describes motion, not the picture. fal calls the prompt on Seedance 2.5 the motion and action you want, and on Omni Flash how the first image should be animated. Kling says the scene is already in the picture, so the prompt needs only the subjects in it and what they do. The rest of its advice comes from a November 2025 guide for older Kling versions and the September 2026 how-to:

• Subject plus movement. Say who or what moves, then how, with one main action and at most one supporting movement. A bare instruction with no subject can leave a photo static or turn a painting into a pan.

• Plain words, plausible movement. Describe movement that could happen in the picture. A description far from the picture can bring a camera cut or a transition.

• A visible action for a still subject. If the subject barely moves, name it and give it something to do, such as turning, walking or reaching.

• Hold the frame still. Describe the one detail that should change, and if the whole frame should not move, say the camera and the background stay still.

• Choose the camera. Fixed, a pan or a move closer.

• Complex physics is hard. The older guide calls complex physical movement hard at that stage and gives a high throw as an example.

• One change per try. Change one thing at a time when you run it again.

Put together, a prompt might read: The woman turns her head toward the window and smiles. Slow push in. The background stays still. That is our own example of Kling's formula, not a tested prompt.

AIGE adds two helpers. Leave What happens empty in the editor's form and it sends a plain default line, the same text the field shows as its placeholder, which leaves the motion to the model. The smaller Video widget on the board will not make a clip without words. Write it well turns a short idea into 50 to 90 words on the subject, action, camera move, setting, light and mood. It spends no credits but has a usage limit, and Undo the writing brings your words back. It reads only your words and never sees the picture, so check its paragraph against the picture and cut anything that contradicts it.

Limits set by the models and by AIGE

Length. Each model has its own list of whole seconds. The longest single clip is 10 seconds on Omni Flash, 15 on Kling V3, Kling 3 Turbo, Kling o3, Seedance and H3 Max, and 30 on Seedance 2.5. How long an AI video can be has the full table. fal's H3 Max page adds that a finished clip can run up to about 0.7 seconds longer than you ask for.

Shape. On some models the picture decides the shape of the clip, and on others a setting does. This is how AIGE's code treats a start picture:

Where the picture decides, make it in the shape you want the clip. On the timeline, a clip whose shape differs from the timeline's fills the frame and gets cropped, unless you set that clip to Fit. What is an aspect ratio explains the shapes, and what is Seedance covers the two Seedance versions.

Sound on Kling V3. The Video widget and the editor's Make tab have no sound switch for it. A clip is asked for with sound and the price includes it, so muting the clip afterward in the editor does not bring the credits back. A plan that reaches only silent Kling V3 makes it without audio and says so under the button.

The end picture. Kling's start-and-end-frames guide, from November 2025, says to pick two pictures that are alike, because large differences can cause a shot switch, a cut to another view, and to keep the transition within 5 seconds. Kling 3 Turbo's flow sends only the start picture. The Video widget drops it from the model list once you add an end picture; the editor's From a picture form keeps it in the list, so choose another model before you add one.

What one clip costs

Take Kling V3, the model the Video widget lists first, at the Canvas defaults: Standard quality, sound on. A 5-second clip costs 63 credits. The same settings at 10 seconds cost 126 credits. The optional Change it first step, which edits the start picture from a line of words, is charged on its own at 8 credits, and every run is priced and charged separately.

Other models price by their own length and quality, and plans differ in reach. In AIGE's default plan table, which can change, Starter makes Kling V3 clips of 3 to 5 seconds, and when your plan cannot make a model or a length, the button names the plan that can. Trust the button over any table. AI video for YouTube has the per-model table for a 6-second clip, Kling image to video, step by step prices the three Kling versions side by side, the Seedance page covers that family, and the pricing page lists the plans.

When to copy a movement instead

From a picture is for motion that words can carry. Copy a movement is for motion you already have on video: a dance, a gesture, a walk. You give it a clip and a picture, and the character in the picture performs the clip's movement. Kling's own animate-a-picture how-to points to a separate motion-control workflow for a character action that has to follow a specific performance.

The differences are practical. There is no length slider: the result is as long as your movement clip, up to 10 seconds when the character faces like the picture and up to 30 when it faces like the clip. It bills in 5-second steps, and in AIGE's default plan table it is off on Starter. Kling Motion Control explained has the rules, and what is Kling AI covers the versions.

Honest note: we ran no test set for this article, so it shows no output and rates none. The picture and wording advice is Kling's, from guides read on 28 October 2026: two dated November 2025, which describe older Kling versions and give lengths that are not today's, and one dated September 2026. Seedance's own prompt guides did not load for us, so nothing here comes from them. The widget, the editor and the sound statement are about the Canvas on a laptop; AIGE's other video controls have their own settings. Vendors change their pages, and AIGE sells the plans, so weigh that.

Quick answers