30 AI Image-to-Video Prompts That Actually Work.

The image-to-video prompts that actually work name one camera move and one subject motion, then stop. The model already sees your photo, so your job is to describe what should change, not re-paint the scene. Below are 30 copy-paste templates grouped by camera move, motion type, and style, plus the formula behind them.

Every prompt here is written to drop straight into an image-to-video tool. Swap the bracketed parts for your own subject and you have a working prompt.

The formula every good prompt follows

Most weak prompts fail for the same reason: they describe the image instead of the animation. The model can already see the picture. What it needs from you is motion. Use this order:

1. Camera move — dolly, pan, orbit, zoom, static, handheld.
2. Subject motion — what moves inside the frame, and how.
3. Speed — slow, gentle, steady, fast.
4. Mood or light — one phrase, optional.

Example: "Slow dolly-in toward the subject, steam rising from the coffee cup, gentle pace, soft morning light." Four parts, one sentence. That is the whole trick.

Prompts by camera move

Camera moves do most of the heavy lifting. They add life even when the subject barely changes, which keeps faces and hands stable. Start here for portraits and products.

• Slow dolly-in toward the subject, everything else still, cinematic and steady.
• Slow dolly-out revealing the surroundings, calm pace, even light.
• Smooth left-to-right pan across the scene, gentle and continuous.
• 360-degree orbit around the subject, slow, keeping the subject centered.
• Subtle push-in with a slight handheld shake, documentary feel.
• Gentle crane-up rising above the subject, slow reveal of the background.
• Locked-off static camera, only the subject moves, no drift.
• Slow tilt-up from the ground to the sky, steady and smooth.
• Rack focus from the foreground to the subject, shallow depth of field.
• Slight parallax drift, foreground moving faster than the background, subtle 3D feel.

Prompts by motion type

This is where you tell the model what comes alive inside the frame. Keep it to one moving element per clip. Two competing motions are the most common cause of warping.

• Hair and clothing drifting in a light breeze, everything else still.
• Subject slowly turning their head toward the camera and smiling.
• Steam rising from the cup, fabric settling, very gentle ambient motion.
• Water rippling on the surface, slow reflections shifting.
• Leaves and grass swaying softly, wind passing through the frame.
• Subject blinking and breathing naturally, a living portrait.
• Candle flame flickering, warm light dancing on nearby surfaces.
• Fabric or flag rippling in the wind, slow and continuous.
• Crowd in the background moving while the foreground subject stays still.
• Rain falling gently, droplets sliding down a window.

Prompts by style

Style cues shape the feel without changing the content. Pair one of these with a camera move and a motion for a finished prompt.

• Cinematic slow-motion, shallow depth of field, golden-hour light.
• Vintage film look, slight grain and gentle flicker, nostalgic mood.
• Dreamy soft-focus, slow drifting motion, ethereal atmosphere.
• Product hero shot, smooth slow rotation, clean studio lighting.
• Anime-style motion, expressive and fluid, gentle background sway.
• Moody noir, low-key lighting, slow creeping camera, soft shadows.
• Bright and energetic, quick subtle bounce, vibrant colors.
• Documentary handheld, natural light, slight realistic shake.
• Macro close-up, slow reveal of fine texture and detail.
• Epic wide landscape, slow drifting clouds, vast and calm.

Quick reference: which prompt for which shot

Three fixes when a prompt misbehaves

• Face or hands warp— slow the motion down. Add "subtle" or "barely moving" and drop any fast camera move.
• The model redraws your image — you described the scene. Cut everything except the camera move and the motion.
• Nothing moves enough— name a specific element to animate (hair, steam, water) instead of asking for general "movement."

Test the same prompt across models

The same prompt lands differently on Kling, Seedance, Wan, Hailuo, and Veo. One holds your source image almost perfectly; another adds more energy and risk. Rather than guess, getvivix runs them on one subscription and shows the exact credit cost on the Generate button before you click, so a 10-second test never surprises you. Generate the still first with the AI image generator, then animate the frame you like and caption it for vertical export.

Frequently asked

What is the best prompt structure for image-to-video AI?

Name the camera move, then the subject motion, then the speed, then the mood. Example: "Slow dolly-in on the subject, hair drifting in a light breeze, gentle pace, warm evening light." The model already sees your image, so describe what changes, not what is there.

Why does my image-to-video prompt get ignored?

Usually the prompt asks for too much at once or describes things already visible in the still. Pick one camera move and one subject motion per clip. Re-describing the scene wastes the prompt and can make the model redraw your image instead of animating it.

Do these prompts work on every model?

The wording works across Kling, Seedance, Wan, Hailuo, and Veo, but each model interprets motion differently. Kling holds the source image tightly, Seedance adds more energy. Running the same prompt on two or three models is the fastest way to find the one that fits your shot.

How long should an image-to-video prompt be?

One or two sentences. Image-to-video models reward precision over length. A short prompt that names a single camera move and a single motion beats a paragraph that lists five things the model has to juggle.

Can I control how fast the motion is?

Yes. Words like "slow," "gentle," "subtle," and "barely moving" reduce warping on faces and hands. "Fast," "rapid," and "whip" add energy but raise the risk of distortion. Start slow for portraits and products, speed up for action shots.

Try these prompts on getvivix — animate any photo across 100+ models, with the credit cost shown before every generation. The free tier gives you 30 credits on signup, no card.