Prompt library
AI camera movements: 46 prompts with previews
AI camera movements are named moves you append to a scene prompt, and naming one is the difference between a shot that drifts and a shot that goes where you meant it to.
Pan, tilt, zoom, dolly, truck, pedestal, slider, arc, orbit, crane, drone, handheld and tracking are the terms Kling, Seedance, Runway, Veo, Hailuo and Pika all answer to, because the footage behind every one of those models was described in exactly those words. This page holds 46 of them across 7 groups. Each card carries the prompt fragment to paste onto the end of your scene description and a looping preview of what the move does, so you can pick the shot before you spend a generation finding out.
About the previews
The prompt wording on this page is mirrored from the freely shared camera movement library at aicameramovements.com. The preview loops are the reference clips published with that library, re-encoded to 480p and served from this domain. None of them were generated by us. The guidance under each prompt is our own.
Jump to a group
How to use these prompts
Write the scene first
Subject, one action, lens and light, duration. The camera clause is a modifier, so it needs something to modify. A camera move on its own gives you a move through nothing in particular.
Append one movement, not two
Paste a single fragment onto the end of the scene. Two moves in one prompt is the most common reason the output comes back with neither, because the model averages them instead of choosing.
Match the move to the subject motion
A tracking shot needs someone to track. A locked static shot needs something in frame that moves on its own. When the camera and the subject want different things, the subject usually wins and the camera clause gets quietly dropped.
Run it once to learn the amount
Every model applies its own distance and speed to the same words. The first run on a new model tells you whether its dolly in travels a metre or a room, and after that you adjust with words like slight, half a step or all the way.
Join moves in the edit
A crane down followed by an orbit is two generations and one cut, not one prompt. The whip pans and the pass through moves are useful here because their blur and their foreground give you somewhere to hide the join.
If the scene half of the prompt is the part giving you trouble, the five slot grammar on the prompt to video AI page is where to start, and the image to video page covers the case where a photo has already answered everything except the camera.
Pan/Tilt
The camera stays where it is and only rotates. These are the cheapest moves to get right, because nothing in the scene changes its distance from the lens.
Static shot
Use it when the subject is doing the work and any camera drift would compete with it. Video models add a slow push by default, so a locked frame has to be asked for in words.
Camera: locked-off static shot. Movement: hold one fixed camera position for the full clip. Speed: still and steady. Framing: keep the same angle, height, lens distance and composition. End: finish with the same framing and camera position.
Pan right
Good for revealing something already in the scene but out of frame, like a second character or the far end of a room. The camera position never moves, so the depth of the shot stays exactly where it started.
Camera: pan right. Movement: rotate the camera horizontally from left to right from one fixed point. Speed: smooth constant rotation. Framing: keep the horizon level while new space enters from the right side of the frame. End: settle on a clear final composition.
Pan left
The same reveal in the opposite direction. Pick the direction your subject is looking or walking, because panning against the eyeline reads as tension whether you meant it or not.
Camera: pan left. Movement: rotate the camera horizontally from right to left from one fixed point. Speed: smooth constant rotation. Framing: keep the horizon level while new space enters from the left side of the frame. End: settle on a clear final composition.
Whip pan right
A cut you can hide inside. The blur in the middle gives you a join between two separate generations, which is the cheapest way to stitch clips without a visible edit.
Camera: whip pan right. Movement: rotate rapidly from the starting direction toward a new target on the right. Speed: fast snap with brief motion blur during the rotation. Framing: begin on one readable composition and land on a second readable target. End: settle into a sharp final frame.
Whip pan left
Use it for reactions and for comedy timing. Because the middle of the clip is deliberately blurred, it also forgives the frame quality these models are worst at.
Camera: whip pan left. Movement: rotate rapidly from the starting direction toward a new target on the left. Speed: fast snap with brief motion blur during the rotation. Framing: begin on one readable composition and land on a second readable target. End: settle into a sharp final frame.
Tilt up
The standard way to make something feel tall: a building, a tree, a person seen from the floor. Start on a detail and finish on the whole thing, not the other way round.
Camera: tilt up. Movement: rotate the camera upward from one fixed point. Speed: smooth constant tilt. Framing: keep the vertical subject or architecture centered as the frame travels upward. End: land on the upper target.
Tilt down
An establishing shot in reverse, going from sky or ceiling down to the thing that matters. It is also the safest way to introduce a subject that is on the ground.
Camera: tilt down. Movement: rotate the camera downward from one fixed point. Speed: smooth constant tilt. Framing: keep the vertical subject or architecture centered as the frame travels downward. End: land on the lower target.
Zoom/Lens
The lens changes and the camera body does not move. That keeps the background fixed relative to the subject, which reads flatter than a dolly but breaks far less often.
Slow zoom in
Builds pressure without moving the camera through the scene, so backgrounds stay stable and artefacts stay hidden. Use it on faces and on anything you want the viewer to look harder at.
Camera: slow zoom in. Movement: slowly increase lens focal length toward a tighter frame. Speed: gradual and even. Framing: keep the main visual target readable as it becomes larger in frame. End: finish on a stable tighter composition.
Slow zoom out
Reveals the context the tight frame was hiding. Best when the surprise is where the subject is rather than what the subject is doing.
Camera: slow zoom out. Movement: slowly decrease lens focal length toward a wider frame. Speed: gradual and even. Framing: keep the main visual target readable as more surrounding space appears. End: finish on a stable wider composition.
Fast zoom in
Punchy enough for a hook and short enough to sit in the first second of a short. Keep the target dead centre or the model loses it partway through.
Camera: fast zoom in. Movement: quickly increase lens focal length toward the main visual target. Speed: quick decisive zoom. Framing: keep the target centered or clearly readable during the scale change. End: finish on a stable tighter composition.
Fast zoom out
Use it for a joke or a scale reveal where timing matters more than elegance. It reads as energetic rather than cinematic, which is the whole point.
Camera: fast zoom out. Movement: quickly decrease lens focal length away from the main visual target. Speed: quick decisive zoom. Framing: keep the target readable as the surrounding space appears. End: finish on a stable wider composition.
Crash zoom in
The loudest move on this page. One per video at most, usually on a reaction or on the detail the voiceover just named.
Camera: crash zoom in. Movement: snap the lens rapidly toward the main visual target. Speed: very fast and punchy. Framing: keep the target readable through the sudden scale change. End: land on a bold tighter composition.
Crash zoom out
The same energy going the other way. It works when whatever appears around the subject is the punchline.
Camera: crash zoom out. Movement: snap the lens rapidly away from the main visual target. Speed: very fast and punchy. Framing: keep the target readable as the surrounding space appears. End: land on a bold wider composition.
Dolly/Track
The camera travels. Dolly moves toward or away from the subject, tracking moves along with it, and both change how foreground and background sit against each other as they go.
Dolly in
Different from a zoom: the camera actually travels, so the background changes size relative to the subject and the shot gains real depth. Ask for this when a zoom came back looking flat.
Camera: dolly in. Movement: move the camera physically forward in a straight line toward the main subject. Speed: smooth controlled push. Framing: keep camera height, lens direction and subject position consistent while distance closes. End: finish in a tighter composition.
Dolly out
Pulls the viewer back out of a scene, which is why it closes a sequence well. State how far you want it to travel or the model keeps going until the subject is tiny.
Camera: dolly out. Movement: move the camera physically backward in a straight line away from the main subject. Speed: smooth controlled retreat. Framing: keep lens direction and camera height consistent while more environment enters frame. End: finish in a wider composition.
Tracking shot
The workhorse for anything with a walking subject. Naming the pace of the subject is what stops the camera drifting off them halfway through the clip.
Camera: tracking shot. Movement: move through the scene with the main subject. Speed: match the subject's pace. Framing: keep the subject consistently readable while the environment moves around them. End: maintain a clear moving composition.
Follow shot / over-the-shoulder
Puts the viewer behind the character so the route ahead becomes the story. Useful when you want a body in frame without committing the model to a face.
Camera: follow shot from behind. Movement: move behind the subject along their route at shoulder height. Speed: match the subject's pace. Framing: keep the back, shoulder or head as the foreground guide while the route ahead stays readable. End: continue following with the subject leading the frame.
Reverse tracking / walk-and-talk
The interview and the walk and talk in one move. It is the only tracking option that holds a face front on for the whole clip, so use it when expression or dialogue carries the shot.
Camera: reverse tracking shot. Movement: move backward in front of the walking subject. Speed: match the subject's forward pace. Framing: keep front-facing face and body framing stable as the background moves behind them. End: hold a clear front-facing moving composition.
Side tracking
Reads as clean and deliberate, and the constant subject distance makes it one of the more stable moves for these models. Good for runners, cars in profile, and product moving past on a line.
Camera: side tracking shot. Movement: move parallel beside the subject along their direction of travel. Speed: match the subject's motion. Framing: keep the subject in side profile or three-quarter profile at a stable distance. End: continue the parallel movement with clear horizontal motion.
Low tracking
Ground level makes ordinary movement look fast and heavy. Feet, wheels, paws and water all gain weight from it.
Camera: low tracking shot. Movement: move at ground or below-waist height alongside the subject's movement path. Speed: match the subject, footsteps or wheels. Framing: keep the low detail readable while the ground plane moves through frame. End: finish with the low perspective clearly maintained.
Vehicle tracking
Ask for this rather than a generic tracking shot when a car, bike or boat is the subject. It tells the model the camera is travelling with the vehicle instead of watching it go past.
Camera: vehicle tracking shot. Movement: move with the vehicle along its route. Speed: match the vehicle's pace. Framing: keep the vehicle stable in frame while the road or environment moves past. End: maintain a clear moving vehicle composition.
Chase shot
Deliberately looser than the other tracking moves. Reframing and small errors read as urgency here, so it forgives the artefacts a steadier shot would expose.
Camera: chase shot. Movement: follow a moving subject quickly along the action route. Speed: fast, reactive and physically close. Framing: keep the subject visible while allowing energetic reframing. End: stay connected to the subject in motion.
Physical Moves
The camera body moves in a straight line or on a curve without following anyone. Short versions of these are the safest way to put motion into a product shot or a still scene.
Truck right
A sideways slide with the lens still pointing the same way. Use it to travel along a wall, a shelf or a line of people without turning the camera at all.
Camera: truck right. Movement: move the camera physically to the right on a straight horizontal path. Speed: smooth constant lateral travel. Framing: keep the lens facing the same direction while the scene slides across frame. End: finish on a clean lateral composition.
Truck left
The same move in the other direction. Trucking left across a scene the subject is walking into makes the space feel larger than a pan of the same length would.
Camera: truck left. Movement: move the camera physically to the left on a straight horizontal path. Speed: smooth constant lateral travel. Framing: keep the lens facing the same direction while the scene slides across frame. End: finish on a clean lateral composition.
Pedestal up
The camera rises but keeps looking straight ahead, so the angle on the subject never changes. This is the move for going from feet to face without a tilt.
Camera: pedestal up. Movement: move the entire camera vertically upward in a straight line. Speed: smooth constant lift. Framing: keep the lens level and pointed in the same direction during the vertical move. End: finish with the higher framing clearly readable.
Pedestal down
Drops the eye line while keeping the lens level. Useful for landing on a table top or a floor detail at the end of a shot.
Camera: pedestal down. Movement: move the entire camera vertically downward in a straight line. Speed: smooth constant descent. Framing: keep the lens level and pointed in the same direction during the vertical move. End: finish with the lower framing clearly readable.
Slider right
A short slide, not a journey. The value is the parallax between foreground and background, which is what makes a still scene look like it was filmed rather than generated.
Camera: slider right. Movement: slide the camera a small distance to the right. Speed: slow controlled constant motion. Framing: keep foreground, subject and background layers readable as parallax shifts. End: finish on a refined composition with the new right-side angle visible.
Slider left
The best default move for a static object such as a product on a table. Short distances also mean fewer frames for the model to get wrong.
Camera: slider left. Movement: slide the camera a small distance to the left. Speed: slow controlled constant motion. Framing: keep foreground, subject and background layers readable as parallax shifts. End: finish on a refined composition with the new left-side angle visible.
Push past / pass-by shot
Passing a foreground object gives a shot instant depth and doubles as a wipe you can cut on. Doorways, pillars, branches and shoulders all do the job.
Camera: push past. Movement: move forward past a visible foreground object, edge or opening. Speed: smooth forward glide. Framing: let the foreground pass close to the lens while the space beyond becomes clearer. End: arrive inside or beyond the foreground layer.
Arc right
A partial orbit that changes the angle without going all the way round. Use it when you want a new view of a subject but not a full turntable.
Camera: arc right. Movement: move on a shallow curved path around the main subject toward the right side. Speed: smooth measured curve. Framing: keep distance, height and subject readability consistent while the angle changes. End: finish from a new right-side angle.
Arc left
The mirror of arc right, and worth alternating with it across a sequence so two consecutive shots do not feel like the same take.
Camera: arc left. Movement: move on a shallow curved path around the main subject toward the left side. Speed: smooth measured curve. Framing: keep distance, height and subject readability consistent while the angle changes. End: finish from a new left-side angle.
Orbit clockwise
The reliable product shot. A centred object on a plain background survives a full circle far better than a person or a busy street will.
Camera: clockwise orbit. Movement: circle clockwise around the main subject at a consistent radius. Speed: smooth controlled orbit. Framing: keep the subject centered while the background rotates around them. End: complete the intended arc or full circle with stable framing.
Orbit counterclockwise
Identical mechanics, useful mainly so a second product shot in the same video does not repeat the first one exactly.
Camera: counterclockwise orbit. Movement: circle counterclockwise around the main subject at a consistent radius. Speed: smooth controlled orbit. Framing: keep the subject centered while the background rotates around them. End: complete the intended arc or full circle with stable framing.
Drone/Crane
Height plus travel. Use these when the point of the shot is the size of the place rather than the detail of the subject.
Crane up
Rises through open air with the subject shrinking below, which is the classic closing shot. Give it something to rise toward or it just gets further away.
Camera: crane up. Movement: travel smoothly upward through open space. Speed: slow controlled vertical lift. Framing: keep the subject or location readable as the camera rises. End: finish with the higher scale clearly visible.
Crane down
Comes down out of open space onto the subject, which makes a good opening. It sets the location before it sets the character.
Camera: crane down. Movement: travel smoothly downward through open space. Speed: slow controlled vertical descent. Framing: keep the subject or location readable as the camera descends. End: finish with the lower subject or destination clearly visible.
Drone push in
Covers ground quickly and suits landscapes, rooftops and water. Name the destination so the flight has a target rather than only a direction.
Camera: drone push in. Movement: fly smoothly forward through open space toward the subject or destination. Speed: controlled aerial glide. Framing: keep the route and destination readable as the camera approaches. End: arrive at a closer aerial composition.
Drone pull back
The reveal: start on a subject and finish on the whole valley. It falls flat when there is nothing worth revealing, so only use it where the wide frame earns its seconds.
Camera: drone pull back. Movement: fly smoothly backward away from the subject or destination. Speed: controlled aerial retreat. Framing: keep the subject readable as more landscape appears. End: finish on a wider aerial composition.
Helicopter shot
Higher and slower than a drone move, and built for scale rather than detail. Good for establishing a city, a coastline or a convoy.
Camera: helicopter-style aerial shot. Movement: move from high altitude along a broad gradual flight path. Speed: steady controlled aerial motion. Framing: keep the landscape or distant moving subject readable at wide scale. End: finish on a stable high-altitude composition.
Human Camera
The camera is attached to a person. Both of these trade stability for the feeling that somebody is holding it.
Handheld shot
Adds the small imperfections that make generated footage read as filmed rather than rendered. Good for documentary and vlog inserts, wrong for product.
Camera: handheld shot. Movement: hold the camera at human operator height with natural body movement. Speed: responsive and organic. Framing: keep the subject readable while the frame has subtle sway and micro-adjustments. End: finish with a natural handheld composition.
Body-mounted camera / Snorricam
The subject stays locked in the middle of frame while the whole world swings behind them. It is a strong, disorienting look, so it belongs on one shot rather than a sequence.
Camera: body-mounted Snorricam. Movement: keep the camera fixed relative to the subject's torso or face while the subject moves. Speed: match the subject's body motion. Framing: keep the subject close, centered and facing the camera as the background moves around them. End: finish with the subject still locked in frame.
Specials
Effects rather than moves. Each one is a look in its own right, so one per video is usually the limit.
First-person view
Hands and arms in frame are what sell it, so say they are there. Suits tutorials, game style footage and anything that should feel like the viewer is doing it.
Camera: first-person view. Movement: move forward at human eye height from the character's perspective. Speed: natural walking or reaching pace. Framing: use visible hands, arms or body edges as the viewer's physical reference. End: arrive at the next point of action from the same point of view.
Tilt-shift
Turns a real location into a model village by faking a very shallow band of focus. Best from a high angle with plenty of small moving things in the frame.
Camera: tilt-shift miniature view. Movement: hold or glide from a high angled view over the scene. Speed: small precise movement. Framing: keep a narrow band of sharp focus across the key subject area with soft blur above and below. End: finish with the miniature-scale view intact.
Infinite zoom
A zoom that never lands, meant to be chained so one world becomes the next. It is the move behind most looping and endless scroll videos.
Camera: infinite zoom. Movement: zoom continuously inward toward the exact center target. Speed: smooth accelerating zoom. Framing: keep the circular target centered as it expands. End: finish when the next visual world fills the frame.
Earth zoom out
Goes from one spot to planet scale in a single clip. Strong as an opener, and it needs a clearly centred starting point or the scale change loses its anchor.
Camera: earth zoom out. Movement: pull upward from the starting point through street, city, landscape and planet scale. Speed: rapid expanding zoom out. Framing: keep the original location centered as scale grows. End: finish on a planet-scale view with the starting point still implied at center.
Time-lapse
The camera holds still and time does the moving. Skies, crowds, traffic and shadows are what make it work, so put at least one of them in the frame.
Camera: locked-camera time-lapse. Movement: hold one fixed camera position while time moves rapidly forward. Speed: fast time compression with a stable camera. Framing: keep the same composition and horizon as motion passes through the frame. End: finish from the same camera angle with visible passage of time.
Pass-through objects
Moving through a wall, a window or a keyhole gives you a transition no cut can match. Pair it with a second clip that starts inside the space you arrived in.
Camera: pass-through movement. Movement: move forward toward a visible object, surface or barrier and continue into the space beyond. Speed: smooth centered glide. Framing: keep the opening or surface centered as the transition point. End: arrive inside the revealed space beyond.
Picking a move for the job
| What you are shooting | Start with | Why |
|---|---|---|
| A product on a table | Slider left, or a clockwise orbit | Short travel and a plain background give the model the least to get wrong. |
| A face | Slow zoom in, or a dolly in | Both build attention without asking the model to rebuild the face from a new angle. |
| A landscape | Drone pull back, or a crane up | The reveal is the shot, and open air is the easiest thing in the world to animate. |
| Someone walking | Side tracking, or reverse tracking | Constant distance keeps the body stable. Reverse tracking is the one that keeps the face. |
| A hook for a short | Crash zoom in, or a whip pan | Both are loud in under a second, which is all the time a short gets. |
| A transition between two clips | Push past, or pass through objects | The foreground crossing the lens covers the join between two separate generations. |
| An establishing shot | Crane down, or a helicopter shot | Both set the place before they set the subject, which is what establishing means. |
| B-roll under a voiceover | Static shot, or a slow zoom in | Quiet on purpose. A busy camera under a voice pulls attention off the words. |
What these prompts do not fix
- Hands, text and faces at distance. A camera move changes the framing, not the model. If a hand falls apart in a static shot it will fall apart in an orbit too.
- Length. Most of these moves are written for a five to ten second clip. A slow crane over three seconds does not read as a crane, it reads as a wobble.
- Continuity between clips. Two generations with the same camera move are still two different scenes. Repeating the lens and light wording across both is what makes them cut together.
- The subject. A perfect crane up on a boring subject is a boring shot from higher up.
For the rest of the pipeline, from scene count to export, the walkthrough is on the how to make AI videos page, and the overview of what these tools can and cannot do sits on the video AI generator home page.
Camera movement questions
What are AI camera movements?
They are the named camera moves from film language, written into a video prompt so the model animates the frame the way you want instead of picking for itself. Pan, tilt, zoom, dolly, truck, pedestal, arc, orbit, crane, drone and handheld are the main ones, and every current model recognises them because the footage they were trained on was described in those terms.
Do camera movement prompts work in every video model?
The vocabulary transfers, the amount does not. Kling, Seedance, Runway, Veo, Hailuo and Pika all understand dolly in and orbit clockwise, but each one applies a different distance and speed to the same words. Expect to run a move once on a new model to learn how far it goes, then adjust the wording rather than the move.
Can I put two camera movements in one prompt?
You can, and the hit rate drops sharply when you do. Asking for a crane down that becomes an orbit usually gives you a blurry average of both. Generate one move per clip and join them in the edit, which is also how the shot would have been filmed.
Where in the prompt should the camera movement go?
After the subject and the action, before the lens and lighting. The start of a prompt carries the most weight and the subject needs to own it, but anything left to the end of a long prompt tends to get dropped, so the camera move should not be last either.
What is the difference between a zoom and a dolly in AI video?
A zoom changes the lens, so everything scales together and the shot stays flat. A dolly moves the camera through the scene, so the background changes size at a different rate to the subject and the shot gains depth. When a zoom comes back looking like a crop, a dolly is what you actually wanted.
Why does the model ignore the camera movement I asked for?
Usually one of three things: the prompt has two moves in it, the move contradicts the subject motion, or the move is buried at the end of a long prompt. Shortening the prompt and moving the camera clause earlier fixes most of it. Some subjects also override the camera, because a fast moving subject pulls the model toward a tracking shot no matter what you asked for.
Do these prompts work for image to video as well as text to video?
They work better there. An input image has already answered the subject, the lens and the light, so the camera clause is close to the only instruction left and the model gives it more weight. Slider, orbit, dolly in and push past are the moves that get the most out of a single photo.
Which camera movements are the most reliable?
Short moves on a simple subject. A slider across a product, a slow dolly in on a face, a locked static shot and a clockwise orbit on a plain background all survive generation far more often than a chase shot or a pass through. The more the frame has to be rebuilt between the first and last second, the more chances there are for something to fall apart.