# Between the Frames: Motion Prompts for the Kling 4.0 API > Published 2026-10-28 · https://www.promptzone.com/flaqai/between-the-frames-motion-prompts-for-the-kling-40-api-5eaf A starting image and an ending image can agree about the subject while leaving the movement between them poorly defined. A box begins closed and ends open, but should its lid rotate on a hinge, slide sideways, or disappear before a new scene appears? The images alone do not replace a clear transition brief. For the Kling 4.0 API, use the frames to establish selected visual states and the prompt to explain how the scene should move between them. API means application programming interface: the connection an application uses to submit the images and instructions. The creative task is to make those inputs tell a coherent story. This guide uses an invented walnut keepsake box. The prompts and illustrations are planning examples, not results from a generation test. The same method applies to a product reveal, a change in light, or a short camera approach, but each needs its own movement description. ![Two reference prints show a walnut box closed and open from a matching camera viewpoint.]() Check whether the endpoints agree about the object, setting, and physical action before writing motion. ## Read the endpoints before writing the transition Put the proposed opening and ending images beside each other. List what changed and what should remain steady. For the box, the lid is closed in the opening and raised in the ending. The box base, tabletop, camera angle, and light are intended to remain the same. If the second image also moves the box to another table or changes the camera side, the request now contains several transitions. A prompt that says “open the box naturally” does not resolve which change matters most. Either prepare more compatible source images or deliberately choose the more complex scene change. Check the physical detail that explains the action. For a hinged lid, the visible hinge and the opening direction should make sense in both references. A beautiful ending image with a detached lid gives the model a different visual destination from a lid still attached at the back. Use images you have permission to process. Treat them as selected guidance, not a promise that every surface or small mark will be reproduced throughout the clip. ## Give the prompt the movement the pictures cannot state A useful motion prompt can make three decisions explicit: the action, the camera behavior, and the stable details. For this example: > The walnut box lid rotates slowly upward around its rear hinges until the interior is visible. Keep the box base resting in the same place on the table. Use a fixed camera and steady lighting, without a cut to another view. This describes one action. It does not add unrelated dust, dramatic light changes, a camera orbit, and a second object entering the frame. Those additions may be worthwhile in another scene, but they would make it harder to judge why this transition did or did not work. “Smooth” is a preference, not a mechanical explanation. “The lid rotates around the rear hinges” gives a more specific relationship to inspect. Likewise, “cinematic” does less work here than a clear instruction that the camera should remain fixed. ![A partly opened walnut box rests on a table while a curved ribbon follows the hinge movement.]() Describe the specific movement and the details that should remain steady. ## Add a middle image only when it explains a necessary state An intermediate frame is useful when the route matters, not merely because another image slot is available. A half-open lid can express the intended hinge relationship. Several almost identical copies of the closed box add little explanation while increasing the input set you must keep consistent. The current documented image-to-video request uses a required `image_url`, an optional `image_end_url`, and optional `key_frames`. The opening image, optional ending image, and intermediate images share a maximum of 10 images. With both endpoints supplied, that leaves at most eight intermediate images. Intermediate images are used in array order. Their positions in time are determined by the model. Do not label a middle image “exactly at 2.5 seconds” and then treat that timing as an API control. If a cut must land on a musical beat or a product must be visible for an exact duration, plan to inspect and edit the returned footage separately. The conceptual sequence below therefore shows order, not a measured timeline. The gap between prints does not encode seconds. ![Three box storyboard prints show closed, partly open, and open states with editing scissors nearby.]() Frame order guides the sequence; it does not assign precise timestamps. Choose final pacing in the edit. ## Map the creative brief to supported fields A compact request body for the box could use this documented shape: ```json { "model_name": "kling-v4.0-image-to-video", "prompt": "Raise lid on rear hinges; keep base, camera and light steady.", "image_url": "https://example.com/box-closed.jpg", "image_end_url": "https://example.com/box-open.jpg", "key_frames": ["https://example.com/box-half-open.jpg"], "resolution": "720p", "duration": 8, "aspect_ratio": "16:9" } ``` Replace the example image addresses with suitable input URLs. Here there are three images in total. The eight-second duration is an illustrative choice within the documented range of 3 to 30 whole seconds; it does not assign a time to the half-open frame. The [Flaq AI Kling 4.0 image-to-video API](https://flaq.ai/models/kuaishou/kling-4-0-image-to-video/) is the generation step that accepts this kind of frame guidance and motion prompt. It does not decide whether your endpoint images are a good creative match or carry out the final edit for you. ## Write a different transition for a different kind of change The same three decisions help with other scenes, but the wording should change with the job. For a **camera approach**, describe where the camera moves relative to a stationary subject: “Move slowly closer to the closed box while keeping it on the table.” Check that the ending crop is compatible with that approach. Do not simultaneously request a fixed camera. For a **reveal**, name what moves out of the way and what becomes visible: “The lid opens to reveal the empty interior.” If you add an object inside only in the last image, you are also asking the sequence to explain its appearance. Decide whether that is intentional. For a **light change**, keep subject and camera movement out of the brief unless they are needed: “A soft patch of afternoon light moves across the stationary box.” The ending image should show a lighting destination that supports the requested direction, rather than a different box or room. These are creative instructions to evaluate, not tested recipes or guarantees. They make the reasons for accepting or rejecting a candidate more concrete. ## Review the journey, not only the last picture Watch the full clip before judging the endpoints. In the box scene, inspect whether the base stays supported, the lid remains attached, and the camera stays on the intended side. Look between the supplied states: a convincing final frame does not excuse a distracting transformation halfway through. If the lid takes the wrong route, first check the relationship between the images and the motion instruction. If both are coherent but the result is unsuitable, you can revise the guidance or choose a different creative approach. Save the previous input set so you know what changed. Avoid adding more frames merely to make the request feel more controlled. Pay-as-you-go generation makes a purposeful change more useful than an open-ended series of guesses. “Unlimited” is not an established free-attempt allowance here. Choose a project limit, stop when the sequence serves its purpose, and use editing where exact pacing matters. Disclosure: This guide is published by the Flaq AI team and includes our own service.