Runway Image-to-Video Motion Prompts: Three Starting Examples

Start with one visible action, compare controlled variations, and record what actually changes. These prompt examples are suggestions to test, not measured results.

Share

For a Runway image-to-video experiment, start with one action you want to see and add detail only after reviewing a result. Runway’s Image to Video Prompting Guide says this type of prompt focuses primarily on motion; its Gen-4 guide recommends starting with the essential movement and adding one element at a time. Keep the guidance for your selected model in view. The examples below are original suggestions to test, not tested outputs or promises of camera precision.

Write a small, observable motion brief

Before generating, write down what would make one take usable. For a product shot, that might be keeping the label legible while a hand turns the object. For a portrait, it might be a small head movement with an expression you like.

  1. Choose a source image you are entitled to use and identify the main subject.
  2. Describe one visible action in a short sentence.
  3. If camera movement matters, add a single camera direction to a separate variation.
  4. Keep a copy of each prompt and the model and settings you actually used.

A useful planning pattern is “subject action; camera direction; optional scene movement.” This is an editable writing aid for this article, not an official syntax or named technical protocol. Omit any part you do not need.

Three untested starting examples

Creative aimSuggested promptWhat to inspect
Quiet portraitThe person turns their head slightly toward the window and pauses. Locked camera.Check the face, the size of the movement and whether the final expression suits the scene.
Product detailA hand slowly rotates the bottle a small amount on the table. The camera moves closer.Look for label changes, unstable fingers and movement that hides the product.
Walking shotThe person walks along the path. The camera tracks beside them.Review foot placement, background continuity and the space around the subject.

Adjust these descriptions to your actual image. They are creative prompts to try, not evidence that a particular model will follow each instruction. Keep the first attempt small enough that you can identify the change you requested.

Compare variations without guessing at the cause

If a take is not usable, record the visible problem before changing the prompt. “The label changes after the turn” is more useful than assuming a particular internal setting caused it. Do not rely on an undocumented motion-scale slider, masking feature or camera-control panel.

  • The subject barely moves: try a simpler, more visible action and compare it with the first prompt.
  • The result moves too much: remove one requested movement and inspect a new take.
  • The image changes in an unwanted way: reduce the scope of the experiment or choose a different source image.
  • No variation is usable: stop at your planned attempt limit and consider a different shot or an ordinary editing approach.

These are troubleshooting experiments, not diagnoses of the model’s internal mechanics. Keep the source image and other available settings unchanged where practical, and note any setting the interface does not let you hold fixed. Use a simple record: prompt version, changed element, visible issue, keep or reject, next experiment.

Bring the selected take back to the story

Judge a successful take in context: does it communicate the intended action, fit the surrounding shots and leave room for the edit? A technically striking movement may still be the wrong shot. For work in Montagea, carry over the creative brief and review criteria for the tools and models actually available there. This guide does not imply that Runway is integrated into Montagea.