Image-to-Video Prompts: How to Animate a Photo With AI
How to write image-to-video prompts for Seedance, Veo, Kling and Runway — what to describe, what to skip, and 5 templates for products, people and places.
- Published
- Reading time
- 2 min
Image-to-video is the easiest way to get consistent AI video: you start from a photo you already like, and the model only has to add motion. But the prompt works differently from text-to-video. The picture already shows what is there. Your prompt should describe what moves.
The one rule: describe motion, not the picture
The most common mistake is re-describing the image: "a red bottle on a marble table in a bright kitchen". The model can already see that. Repeating it adds nothing and sometimes makes the model change things.
Instead, describe three layers of motion:
- Camera motion: push-in, pull-back, slow orbit, pan left, static.
- Subject motion: what the person or object does.
- Environment motion: steam, wind, light changing, water, background people.
A simple structure
Camera:
Subject:
Environment:
Mood and pace:
Keep the same:
Avoid: morphing, extra people, distorted hands, text, sudden cuts.
5 templates
Product: slow hero orbit
Camera: slow orbit around the product at the same height, about 30 degrees, smooth and stable.
Subject: the product stays still; light reflections glide across its surface.
Environment: subtle dust particles in the light.
Keep the same: the product's shape, label and colours.
Avoid: warping, text changes, extra objects, camera shake.
Portrait: natural moment
Camera: static, slight natural handheld feel.
Subject: the person blinks, breathes, and gives a small natural smile; hair moves slightly.
Environment: background softly out of focus, gentle light change.
Keep the same: face, hairstyle and clothes.
Avoid: talking, big head movements, face morphing.
Food: steam and texture
Camera: very slow push-in toward the dish.
Subject: steam rises from the food; sauce glistens.
Environment: warm light, slight movement of a napkin edge.
Avoid: hands, utensils appearing, changes to the plate.
Landscape: living postcard
Camera: slow pan from left to right.
Subject: clouds drift; water ripples; grass moves in the wind.
Environment: light shifts slowly as if the sun is moving behind clouds.
Avoid: people, birds flying into frame, sudden lighting changes.
Before and after reveal
Camera: static.
Subject: a soft wipe of light moves across the frame from left to right, revealing the finished .
Keep the same: composition and camera angle.
Avoid: morphing shapes, text, extra objects.
Tips for better results
- Start with a strong image. Sharp, well-lit photos with a clear subject animate best. Our studio product shot prompt is a good way to create one.
- Less motion is more. Big moves cause distortion. Slow camera moves and small subject motion look the most real.
- Short clips first. Test with the shortest duration, then extend the take that works.
- Generate several takes. Results vary between runs; pick the best and move on.
For full text-to-video prompts with timed beats, see our cinematic product reveal and AI video prompts.
Frequently asked questions
How do I write an image-to-video prompt?
Describe the camera move, what the subject does and what moves in the environment, then list what must stay the same and what to avoid. Do not re-describe what is already in the image.
Which tools support image-to-video?
Many AI video tools, including Seedance, Veo, Kling and Runway, can start from an image. Features change often, so check the tool you use.
Why does my AI video distort the subject?
Large or fast movements often cause distortion. Use slower camera moves, smaller subject motion and add an avoid line such as "no morphing".