AI Video Camera Movements: A Prompt Vocabulary With Practical Examples

“Cinematic camera movement” is one of the least useful phrases in an AI video prompt. It asks for a feeling without defining the change in viewpoint that should create it.

A camera move has geometry. It can rotate from a fixed position, translate through space, change focal length, rise, fall, orbit, or inherit small human motion. Naming the geometry helps a video model decide which parts of the scene should reveal, hide, and move relative to one another.

This guide is a practical vocabulary for prompting those moves. It focuses on what the viewer should see, when each move is appropriate, and how to keep camera motion from fighting subject motion.

First distinguish framing, position, and movement

Three prompt elements are often mixed together:

• Framing describes how much of the subject is visible: extreme close-up, close-up, medium shot, full shot, wide shot.
• Position describes where the camera is: eye level, low angle, overhead, profile, three-quarter view.
• Movement describes how that position or view changes over time: pan, tilt, push, truck, crane, orbit.

“Low-angle close-up” is not a move. It is a starting composition. “The camera slowly pushes in from a low-angle medium shot to a close-up” supplies a path and an ending composition.

Start prompts with a stable composition:

Medium-wide shot at counter height, looking across a bakery worktable.

Then add one movement:

The camera slowly trucks right by one meter, preserving its height and distance from the table.

Finally state the reason the move is visually readable:

Foreground flour bags create gentle parallax while the baker remains centered.

Pan

A pan rotates the camera left or right from a fixed position. It does not slide sideways.

Use it to follow lateral action or reveal something beside the initial frame.

Eye-level medium shot of a cyclist moving from left to right. The camera pans right smoothly to follow at a constant distance. The background passes behind the cyclist with natural motion blur; no zoom or forward movement.

A pan can become unstable when the prompt also asks the subject to approach the lens. Keep the subject at a roughly constant distance or choose a different move.

For a reveal:

Begin on the empty workbench, then pan slowly right to reveal the finished wooden chair. End centered on the chair and hold for two seconds.

The beginning and ending targets stop the pan from continuing indefinitely.

Tilt

A tilt rotates the camera up or down without changing its physical position.

Use it to reveal height:

Start on hiking boots standing in shallow snow. Tilt upward slowly along the red jacket to the climber's face, ending in a medium portrait. Fixed camera position, no zoom, steady movement.

A tilt is different from a crane move. During a tilt, perspective stays anchored at the same camera point. During a crane or pedestal move, the camera itself changes height.

Push in and pull back

A push in moves the camera physically closer. A pull back moves it away. Both change perspective and parallax.

Wide shot of a quiet reading room. The camera makes a slow, straight push toward the open book on the central desk, ending in a medium close-up. Shelves move outward at the edges with natural parallax. No lens zoom.

Pushes work well for directing attention, but large travel can deform faces or products because the model must invent detail. For fragile identity, make the move shallow and define the ending frame.

A pull back reveals context:

Begin with a close-up of a paintbrush touching blue pigment. Slowly pull back to reveal the artist, table, and sunlit studio, ending in a wide shot. The brush action remains gentle and continuous.

Do not combine a dramatic pull back with several new actions. The environment reveal is already the visual event.

Zoom

A zoom changes focal length while the camera stays in place. Perspective does not shift like a physical push.

Locked camera at eye level. Slowly zoom from a medium shot to a tight close-up of the watch face. Keep the watch centered and the background compressed; no camera translation or orbit.

Many generated clips interpret “zoom” loosely. If the distinction matters, explicitly say “camera remains in the same physical position.” If it does not matter, “gentle push in” often produces more readable depth.

Truck or dolly sideways

A truck moves the camera left or right through space while orientation can remain constant. Foreground and background layers slide at different speeds, producing parallax.

The camera trucks slowly left along the kitchen island at a constant height, keeping the chef centered in profile. Foreground herbs pass close to lens while the windows move more slowly in the background. No pan and no orbit.

This move needs depth layers. A flat wall behind a centered subject provides few cues, so the output may look like a pan or digital slide.

Pedestal, crane, and jib

A pedestal move raises or lowers the camera while keeping its angle relatively stable. A crane or jib can combine vertical and horizontal travel.

Begin at tabletop height beside the assembled model. The camera rises vertically to an overhead view, revealing the full layout. Smooth crane movement, constant lighting, and no subject motion.

These moves reveal previously unseen geometry, so they ask more of the model than a shallow push. Use a simple scene and limit subject action. If the overhead arrangement must be exact, generate or provide a target last frame when supported.

Orbit

An orbit moves around the subject while maintaining attention toward it.

A matte black speaker remains fixed on a stone plinth. The camera makes a slow 30-degree clockwise orbit from front-left to front-right at the same height and radius. Preserve speaker geometry, control placement, and logo. No full rotation and no zoom.

Specify degrees or start/end viewpoints rather than “orbit around.” A full 360-degree move requires the model to invent the unseen back of the subject and reconcile it with the front. That can be useful for exploration but risky for product accuracy.

Use clean backgrounds and strong subject separation. If nearby objects cross the subject during the orbit, their occlusion must be generated consistently.

Handheld and stabilized follow

Handheld motion adds small imperfections in position and rotation.

Documentary-style handheld medium shot following behind a farmer walking through waist-high grass. Subtle footstep bounce and minor horizontal drift, not shaky. The farmer stays centered; natural morning light.

Words like “subtle,” “controlled,” and “not shaky” matter. Without them, the model may exaggerate the effect.

A stabilized follow is smoother:

Gimbal follow shot moving backward at walking speed as the subject approaches. Constant eye-level height and distance, smooth stabilized motion, no orbit.

Following a subject asks the model to coordinate camera speed and body motion. Keep the path straight and the environment readable.

Rack focus and depth effects

Rack focus is a lens operation rather than camera travel. Attention shifts between depth planes.

Locked medium shot. Begin focused on raindrops on the window in the foreground, then slowly rack focus to the person reading at the desk in the background. No camera movement and no lighting change.

The prompt must name both focus targets and their depth relationship. “Cinematic focus change” is too vague.

Shallow depth of field can accompany a move, but it is not movement. Be careful with extremely shallow focus on fast action: it can cause the subject to dissolve into blur.

Aerial, drone, and top-down moves

An aerial move changes a large amount of scene geometry.

High aerial wide shot above a winding coastal road. The camera glides forward following the road's curve while slowly descending. One small car travels in the same direction below. Stable horizon, no rapid banking.

Limit the number of independently moving elements. A forward glide, descent, curved road, traffic, waves, birds, and time-lapse sunset is too much for one short clip.

For a top-down graphic composition:

Direct overhead view of a desk. The camera remains perfectly vertical and makes a slow clockwise rotation of ten degrees. All objects stay fixed and retain their shapes.

“Perfectly vertical” prevents the model from drifting into an oblique angle.

Combining moves without losing control

Two compatible motions can work when one is dominant and the other is slight:

• slow push in plus a small tilt up;
• truck right plus a gentle pan that keeps a moving subject centered;
• crane upward plus a gradual pull back;
• shallow orbit plus a modest push.

State the primary move first and quantify the secondary:

The camera primarily trucks right along the counter while panning slightly left to keep the cup centered.

Avoid contradictory or overloaded combinations:

• locked camera plus orbit;
• pan plus “camera stays pointed straight ahead”;
• zoom in plus pull back unless you intentionally want a dolly-zoom effect;
• full orbit plus rapid subject spin;
• handheld plus perfectly static horizon.

Match camera choice to editorial purpose

Choose the move after deciding what the shot must communicate:

The move should reveal information. Motion added only to prevent a static frame often produces decorative drift that weakens the subject.

Prompt construction and testing

Google's Veo prompt guide treats camera framing, position, movement, lens effects, and visual style as distinct ingredients. Use that separation in your own prompt:

[Framing and position]. [Subject action]. [One camera path with beginning and ending]. [Lens or focus behavior]. [Lighting]. [Preservation constraints]. [Audio and exclusions].

Test the camera with a simple subject before adding complex action. If it fails, shorten the travel, define start and end, remove conflicting style words, and name what must remain centered.

In a multi-shot piece, record the selected framing and movement in the storyboard template. Repeating one movement family—mostly locked shots with two shallow pushes, for example—creates a more coherent visual language than using every technique in this list.

Final checklist

Before generating:

• Is the starting framing explicit?
• Is the camera's physical behavior correctly named?
• Is there only one primary movement?
• Are beginning and ending compositions defined?
• Does the scene contain the depth needed to show the move?
• Is subject movement simple enough to coexist with it?
• Have contradictory instructions been removed?
• Does the movement serve the shot's purpose?

Camera vocabulary is not magic syntax. It is a way to make spatial intent testable. When a result is wrong, you can revise “truck right at constant height” far more intelligently than “make it more cinematic.”