Prompting guides
How to Write Image-to-Video Prompts: 5 Motion Examples
Your photo already contains a scene. The missing instruction is what should happen next. This guide turns that decision into a short motion prompt, with examples for portraits, products, landscapes and action shots.
Quick answer
Start with one visible action, choose one camera movement, and describe how the shot ends. Use an image that can plausibly support that motion. Generate a simple baseline before adding another moving element.
What should an image-to-video prompt describe?
Describe a change that a viewer could point to: a head turns, a curtain lifts, or the camera moves closer. Words such as “beautiful” and “cinematic” may suggest a style, but they do not tell you whether the requested movement happened.
Use this editable planning pattern:
[Visible subject] performs [one action]. The camera [one movement or fixed position]. The action [pace and ending].
For the cover scene, a first test could be: “The hanging raincoat sways gently in the breeze. The camera stays in a fixed wide shot. The movement remains small throughout.” This gives you one thing to evaluate. Add moving grass only after the main action works.
These are starting prompts, not tested outputs or commands that every model follows exactly. Review each result against the source photo.
How to choose a photo that supports the movement
Inspect the parts that need to move before writing the prompt. A cropped wrist gives little information for a waving hand; a front-facing package does not show what its back looks like. Avoid asking the model to reveal crucial details that the photo never shows.
Match the action to what is visible
- For a portrait, begin with a small expression or head movement instead of a full-body walk.
- For a product, keep the visible side in view for your first test.
- For a landscape, pick one moving layer such as grass, water or clouds.
Leave room in the frame
If the subject will move right, allow space on its right. If you want a closer view, check whether the crop can tighten without cutting off the key detail. Select a source crop suited to your intended output shape; a prompt alone cannot recover missing composition.
Which camera movement should you use?
Choose the camera based on what you want the viewer to notice. For an initial test, a fixed camera makes subject movement easier to judge. Introduce a camera move when it serves the shot.
| Direction | Use it to | Check the result for |
|---|---|---|
| Fixed camera | Show a small gesture or moving material | Unrequested drifting or reframing |
| Slow push-in | Bring attention to a face or surface detail | Distortion as the subject fills the frame |
| Gentle sideways slide | Show depth between foreground and background | Objects sliding together like a flat picture |
| Tracking shot | Follow a clearly visible moving subject | Changing distance, direction or body proportions |
A pan rotates the view; a sideways slide changes the camera position. State which you mean instead of combining both terms. A requested move is still a creative instruction, not a precise camera rig.
5 image-to-video prompt examples to adapt
Replace the subject and action with details that are actually present in your photo. Start with one example, then change a single instruction if the result misses your intention.
1. Portrait: a small expression change
The person makes a small, relaxed smile while looking toward the camera. The camera holds its position at eye level. The expression settles gently at the end.
Use a clear face with an unobstructed mouth. Check the eyes, teeth and jaw through the transition, not just the first frame.
2. Product: a restrained detail reveal
The camera moves slowly closer to the front of the bottle. The bottle remains upright on the pedestal. The move ends with the cap and upper surface clearly in view.
Check the silhouette and cap shape. For a complete commercial sequence, use the product-photo video workflow.
3. Landscape: one moving layer
The reeds bend gently in a passing breeze, then ease back. The camera stays fixed on the lake. The distant shoreline holds its position.
Choose a scene with distinct foreground plants. If the lake and horizon also warp, simplify the source or the action before adding more atmosphere.
4. Fabric: subtle material motion
The loose curtain lifts slightly from the open window and settles back. The camera remains at a fixed distance. The movement is light and unhurried.
Leave the curtain edge visible. Look for fabric joining the window frame or changing its pattern.
5. Action: follow a short movement
The cyclist rolls forward along the path at an easy pace. The camera follows from the same side at a steady distance. The cyclist continues in the same direction throughout.
Start from a photo with the bicycle and rider visible. Wheels, pedals and limbs need closer inspection than a static product shot.
How to try a motion prompt in Flow AI Video
- Open the image-to-video generator and upload your source image.
- Write the motion instruction in the prompt field. Keep a copy so you can compare later attempts.
- Review the available model, duration, aspect ratio and credit cost shown in the generator. Options depend on the selected model; choose a short supported duration for the first test.
- Generate one baseline clip. Watch it once for the overall action, then pause to inspect details.
- Keep the image and settings unchanged while revising one part of the prompt. Compare the new result with the baseline.
Do not assume changing models preserves the same settings or cost. Review the controls again before submitting another job. You can plan an ending in the text, but exact timing is not guaranteed.
How to fix unwanted movement, cuts and distortion
Separate an unwanted creative result from a failed generation request. Rewriting the prompt can help with a visible motion problem; it does not repair an upload or service error.
| What you see | Next controlled test | What to compare |
|---|---|---|
| The camera moves when it should stay still | Request a fixed view and name only the subject movement | Position of the same background corner at the start and end |
| The clip changes scenes | Remove multiple locations or events; describe one continuous action | Whether the background remains the same throughout |
| Face or object details change | Reduce the turn or camera angle change | Key features in the first, middle and final frames |
| The result hardly moves | Replace vague mood words with one visible action | Whether that action can be identified without reading the prompt |
If successive revisions repeat the same defect, change the source image or try another suitable model. Keep notes on the image, settings and result so you can tell which change helped.
Image-to-video prompting questions
Should I describe every detail in the photo?
No. Refer to enough detail to identify the subject you mean, then explain the movement. Add appearance constraints only when a specific detail needs attention.
Can one prompt make several finished shots?
That depends on the model and workflow. For a controllable first attempt, generate one shot at a time and assemble approved clips in an editor. Do not expect a long list of events to fit a short clip.
Will a longer prompt improve the result?
Only if the extra words resolve a real ambiguity. Compare a simple baseline with one added instruction, rather than rewriting everything at once.
