Image-to-video: animate a starting image with purposeful motion.

Published by ClapGen · Updated 25 October 2026

Image-to-video uses a still image to communicate an appearance or opening composition. The next decision is what should move, how the viewpoint should behave and which image role the selected workflow supports. The public release currently uses simulated requests. Live generation and paid subscriptions are unavailable.

1. Start with an image suited to the shot

Choose a clear image with the intended subject, framing and light already visible. Check hands, facial features, object edges and small lettering before asking for motion; an existing defect may become more distracting when it moves.

Leave enough room for the planned action. If a hand needs to reach an object or a person needs to walk out of frame, that relationship should make sense in the starting composition.

For a single-character scene, avoid accidentally supplying a contact sheet, a collage or a portrait-and-full-body diptych as though it were one natural shot. Prepare an appropriate single-view composition first.

2. Assign the image to the correct role

In a Frames workflow, the first-frame slot communicates the requested opening composition. A last-frame slot, when offered, communicates an ending composition. General Reference workflows can use an image to guide appearance without promising that the supplied picture is the first frame.

Choose a compatible model before adding media. Some workflows support only a first frame, while others also expose a last frame or intermediate keyframes. The presence of several images in a library does not assign those roles automatically.

If using a last frame, keep its subject and intended transformation compatible with the start. A large unexplained change in pose, setting or perspective asks the model to invent the transition.

Explore Frames and Reference controls in Video Studio

3. Describe movement that the image does not show

Use the prompt to describe the action, environmental movement and camera direction. You can refer to the subject already visible in the image instead of replacing its entire appearance with a new verbal description.

Original example brief for a still image of a wrapped gift: The loose ribbon ends sway gently in a light breeze while the box remains on the table. The camera makes a slow move closer and holds the box in the center of the frame.

This is an illustrative motion brief, not a tested result. It identifies what moves, what stays fixed and how the viewpoint changes. If you want a transformation that introduces something absent from the image, explain that change clearly.

4. Keep motion and framing manageable

For a portrait, begin with a small, coherent action such as a slight turn rather than several simultaneous gestures. For a product scene, decide whether the subject moves, the camera moves or both. A fixed camera can make an object movement easier to inspect.

Choose the duration, format and resolution that the selected workflow actually offers. A source image may determine some settings, so follow the visible controls and validation rather than assuming every format is available.

Changing a face, clothing, camera direction and background in one request makes it difficult to identify why a result drifted. Keep the features that matter stable while changing one creative decision.

5. Inspect continuity across the whole clip

For a future live output, compare the opening with the supplied frame and check later moments for changes in face, hair, clothing, object shape and background detail. An opening-frame request is not a guarantee of a pixel-identical frame or perfect preservation.

If the camera moves but the subject also appears to slide, simplify the motion directions. If appearance changes, check whether the image was placed in a first-frame slot or a general reference role and whether the prompt contradicts it.

Review the ending, not only an attractive thumbnail. Unwanted shape changes and lost objects may appear after the first few moments.

When a still image is not enough

A still image does not specify the timing of a recorded performance. If you need a particular gesture, movement or interaction from an existing clip, investigate a source-video or Motion Control workflow.

If several shots must depict the same fictional character, prepare a consistent set of identity references and an agreed visual brief before making each scene. One image-to-video request does not establish permanent identity across future requests.