Image-to-Video Prompts — 12 Motion Templates That Work (2026)

Q: Why does my image warp or melt in image-to-video?

The prompt asks for too much motion or re-describes the scene. Use one small motion, add anchors like background perfectly still, and prefer slow movement.

Q: Do image-to-video prompts work the same on every model?

The principles transfer, but Kling and Seedance follow motion instructions most literally, Runway responds well to camera terminology, and budget models need shorter prompts.

Subtle life (cinemagraph feel)

Subtle motion only: hair and clothing sway gently in a light breeze, background perfectly still

BREEZE — brings portraits to life without warping the face

The scene stays still while steam rises slowly and light flickers softly

AMBIENT — for food, coffee, candles, neon; one moving element sells the whole shot

Gentle water ripples, reflections shimmer, everything else motionless

WATER — lakes, rain puddles, glasses; the classic living-photo effect

Camera motion (image stays intact)

Slow push-in toward the subject, no scene changes

PUSH-IN — the safest i2v prompt in existence; adds drama to any image

Slow parallax pan to the right, foreground moves faster than background

PARALLAX — fakes 3D depth from a flat image; strongest with layered scenes

Camera slowly orbits around the subject, revealing its side profile

ORBIT — the model invents unseen angles; test on simple subjects first

Slow pull-back revealing more of the environment around the subject

REVEAL — the model extends the scene beyond the frame; great for landscapes

Subject action

The person turns their head slowly toward the camera and smiles slightly

TURN — small, specific, single action; the recipe for faces that don't melt

The character walks forward at a natural pace, camera tracking alongside

WALK — pair a subject action with one camera move, never more

The product rotates slowly 360 degrees on its axis, studio lighting unchanged

PRODUCT SPIN — e-commerce staple; "lighting unchanged" prevents color drift

Atmosphere & drama

Storm clouds roll across the sky in timelapse while the landscape below stays sharp

SKY TIMELAPSE — movement in the sky, stability on the ground; very reliable

Snow falls gently through the scene, accumulating motion in the air only

PARTICLES — snow, rain, dust, embers; adds cinema without touching the subject

The 4 rules of i2v prompting

Describe motion, not the scene. The model can see your image — re-describing it invites the model to redraw it. Prompt only what should change.
One motion at a time. "She turns, waves, and the camera orbits while snow falls" guarantees warping. Pick the single most valuable movement.
Anchor what must not change: "background perfectly still", "lighting unchanged", "face stays sharp" — negative space in your motion description is as important as the motion.
Slow words are stability words: "slowly", "gently", "subtle" reduce artifacts on every model. Fast motion is where i2v breaks.

Works with Kling, Runway, Seedance/Dreamina, Wan and most i2v tools. Building the image first? Get the scene right with the prompt generator, and compare i2v generation costs in the cost calculator.

FAQ

Why does my image warp or melt in image-to-video?

Usually the prompt asks for too much motion, or re-describes the scene so the model redraws it. Use one small motion, add anchor phrases like "background perfectly still", and prefer slow movement — fast action is where i2v models break down.

Do image-to-video prompts work the same on every model?

The principles transfer — motion-only descriptions, single actions, stability anchors — but strictness varies. Kling and Seedance follow motion instructions most literally; Runway responds well to camera terminology; budget models need shorter, simpler prompts.