Some images look like a paused video frame. Others look like a taxidermy diorama. The difference is rarely about the subject — it's about a handful of composition and prompt choices that tell your eye something just happened or is about to. If your generations keep coming out stiff, this walkthrough covers the cues that fix it. You can try any of these directly with the AI image tool on the homepage while you read.
Start With an Unfinished Action
The single biggest reason generated images look frozen is that the prompt describes a completed state. "A dancer standing on a stage" is a pose. "A dancer mid-turn, hair still catching up to her shoulders" is a moment. Language that names a body part lagging behind, a limb at the apex of a swing, or fabric that hasn't settled yet gives the model something to imply.
Try swapping static verbs for ones with implied sequence: catching, landing, turning, reaching, spilling, snapping back. None of these are motion blur — they're motion grammar.
Use Diagonal Lines Instead of Horizons
A perfectly level horizon reads as calm and static. Tilt the composition a few degrees and the whole frame gains energy. In practice this means describing the camera angle rather than the subject: "shot from a low angle, slight tilt" or "camera rolled three degrees." You can also build diagonals into the scene itself — a staircase, a diagonal shadow across pavement, a road running corner to corner.
Vertical and horizontal lines settle an image. Diagonals unsettle it. That's most of the trick.
Directional Blur, Applied Sparingly
Blur sells speed, but the wrong kind flattens a picture into mush. The goal is selective blur: the background streaks while the subject stays sharp, or a single fast-moving element smears while everything else holds. Prompt for it explicitly — "background streaked with horizontal motion blur, subject in sharp focus" — instead of asking for general blur, which tends to soften the whole frame.
Be careful with faces. A blurred face doesn't read as moving, it reads as a mistake. Keep motion blur on edges, background, and secondary elements.
Particles Are Your Cheapest Motion Cue
Dust kicked up, water spray, falling leaves, sparks, steam. Small airborne particles imply that something in the frame recently displaced them, which the brain reads as recent or ongoing movement. Adding one particle system to an otherwise still composition often does more work than a dozen other adjustments.
Keep particles directional. Dust drifting straight down looks like a photograph of dust. Dust trailing behind a runner looks like speed.
Weight and Contrapposto
Bodies at rest distribute weight evenly and look like mannequins. Bodies mid-motion are off-balance. Describe where the weight is: "weight shifted onto the back foot," "leaning forward past the point of balance," "one hand braced against the doorframe." Asymmetry signals that gravity and momentum are actively negotiating, which is exactly what stillness lacks.
Let the Light Move Too
Light that just arrived somewhere is more dynamic than light that's been sitting. Describe light as transient — "late afternoon sun cutting under the awning," "headlights just entering frame from the left," "a window reflection sliding across the wall." Light in transition carries time with it.
What to Avoid
Three habits kill the effect. First, overloading: three motion cues work, eight turn the frame into chaos. Pick one dominant cue and one supporting one. Second, contradictory blur, where the background streaks left and the subject streaks right for no reason. Third, prompting for a pose with a cheerful adjective attached — "energetic pose" tells the model nothing, while "mid-stride, coat flaring behind" tells it everything.
A Repeatable Recipe
Here's a structure worth reusing: subject in mid-action verb, camera angle with a slight tilt, one directional motion cue, one particle or light element in transit, and a specification of what stays sharp. That's it. Five beats, in that order, usually produces a frame you can't quite tell is still.
Give it a run on a few different subjects — a cyclist, a dog shaking off water, someone closing a door — and you'll start to see which cue your eye responds to most. Once you find it, lean on it. Being able to reliably produce images that imply movement is one of the more useful things you can pick up with an AI image generator, and it carries over into thumbnails, banners, and social posts where a flat image loses the scroll.