In current Runway image-to-video guidance, the input image establishes much of the visual information. The text prompt is most useful when it describes how the scene moves over time.
1. Choose a frame that can move
Use a source image with clear subject separation, coherent anatomy and enough space for the intended camera or body motion.
2. Describe motion, not the whole picture again
Write subject motion, environmental motion and camera movement. Avoid repeating colors, clothing and composition already visible unless you need to constrain them.
3. Keep the first shot simple
Ask for one primary subject action and one camera behavior. Too many simultaneous motions increase warping and make failure hard to diagnose.
4. Judge temporal artifacts
Watch hands, face, object edges, background geometry, speed changes and whether motion starts/stops naturally. A good first frame does not guarantee a good clip.
5. Iterate one motion variable at a time
Change camera, subject speed or environmental movement separately. When the base motion works, then build a sequence or edit around it.
The input image defines the subject and composition. Animate only the motion: [subject action]. Camera: [camera move]. Environment: [subtle motion]. Keep identity, clothing and scene geometry stable; natural acceleration and deceleration; no new objects.
Rewriting the image as a long text prompt; requesting orbit + zoom + running + explosion at once; ignoring the last frames; changing the source image and motion prompt simultaneously.
Finish check
The clip preserves the source identity and geometry while one clear subject motion and one clear camera behavior remain stable through the shot.