AI filmmaking and cinematography glossary

A frame so wide that people are small or barely visible, used to show scale, geography or isolation.

In a prompt: "extreme wide shot, a lone rider crossing a salt flat, tiny in frame"

See also: Establishing shot, Wide shot

Shows a person head to toe with room around them, so the audience reads both the body and the space.

See also: Medium shot, Extreme wide shot

Frames a person from roughly the knees or mid-thigh up. Wide enough for gesture, close enough for expression.

Frames a person from the waist up. The default size for conversation: close enough to read a face, wide enough to see hands and gesture.

See also: Medium close-up, Wide shot

Frames a person from the chest or shoulders up. Common for dialogue where the face matters but the body still reads.

The face fills most of the frame. Close-ups carry emotion, and they are where inconsistent character details show first.

In a prompt: "close-up of her face, eyes glistening, soft window light from the left"

See also: Extreme close-up, Character reference

A single detail fills the frame: an eye, a hand on a trigger, a match being struck. Used for emphasis and tension.

A close shot of an object or detail cut into a scene to show information: a phone screen, a letter, a clock.

See also: Cutaway

Two people share the frame, usually at medium size. It establishes their relationship before a scene cuts to singles.

See also: Over-the-shoulder shot, Shot–reverse shot

The opening wide of a scene that tells the audience where and when they are before the scene moves in closer.

See also: Extreme wide shot

The camera sits at the subject’s eye height. Neutral and observational, it is what most models default to when no angle is given.

The camera looks up at the subject, which makes them look larger, stronger or more threatening.

In a prompt: "low-angle shot looking up at the general on the balcony, overcast sky behind"

See also: High-angle shot, Worm’s-eye view

The camera looks down at the subject, which makes them look smaller, weaker or more exposed.

See also: Low-angle shot, Bird’s-eye view

Looking straight down from above. It turns a scene into a pattern or map and removes the horizon entirely.

An extreme low angle from ground level looking up, exaggerating height and scale.

The camera is rolled so the horizon runs diagonally. It signals unease, disorientation or a world off balance.

Framed from behind one character, with their shoulder in the foreground, looking at the person they are facing.

See also: Shot–reverse shot, Two-shot

The camera shows what a character sees, as if through their eyes. Usually cut between a shot of the character looking and a shot of what they look at.

The subject is seen side-on, facing across the frame. Useful for silhouettes and for showing someone lost in thought.

The camera does not move. Static shots let performance and composition do the work, and leave a model less room to invent motion.

In a prompt: "static shot, locked-off camera" stops most models from adding a drift you did not ask for.

The camera rotates left or right on a fixed point, like turning your head. It reveals space or follows action across the frame.

See also: Tilt, Tracking shot

The camera rotates up or down on a fixed point, like nodding. Often used to reveal height, a building or a character from feet to face.

The whole camera moves toward or away from the subject on a wheeled dolly. A push in builds intensity; a pull out reveals context or leaves a character alone.

In a prompt: "slow dolly in toward his face as he realises" reads more clearly to a model than "zoom".

See also: Zoom, Dolly zoom

The focal length changes during the shot so the image magnifies or widens while the camera stays put. Unlike a dolly, perspective does not change.

See also: Dolly shot, Focal length

The camera dollies one way while zooming the other, keeping the subject the same size as the background stretches or collapses. A signal of shock or vertigo.

The camera travels alongside, behind or ahead of a moving subject, keeping pace with them.

See also: Dolly shot, Steadicam shot

The camera rises or descends through space on a crane or jib, often to open or close a scene.

See also: Drone shot

An aerial move flown by a drone: flyovers, rising reveals and top-down passes over landscapes or cities.

The operator holds the camera, so the frame carries small, human movement. It reads as urgent, documentary or intimate.

In a prompt: "handheld camera, slight shake" works on most models; "shaky cam" tends to overdo it.

A stabilised moving shot that glides with the operator, smoother than handheld and freer than a dolly on track.

The camera circles around the subject while keeping them centred. Popular in AI video, and a quick test of whether a model keeps a character stable from every side.

A pan so fast the image smears into motion blur. Often used as a transition by hiding the cut inside the blur.

Focus shifts during the shot from one subject to another at a different distance, steering the audience’s attention without a cut.

See also: Depth of field

Measured in millimetres, it sets how wide or tight a lens sees. Short focal lengths (14–24mm) are wide; long ones (85mm and up) are telephoto.

In a prompt: Naming a lens ("shot on a 35mm lens") is shorthand for field of view and perspective, and most video models respond to it.

See also: Wide-angle lens, Telephoto lens

A short focal length that takes in a lot of the scene, exaggerates depth and distorts faces close to the camera.

Around 35–50mm on full frame, close to how the human eye sees perspective. A neutral, natural look.

A long focal length that magnifies distant subjects and compresses space, so background and foreground seem stacked together.

See also: Lens compression

The flattened look of a telephoto shot, where far objects appear close behind the subject. It comes from camera distance, which long lenses force.

A lens that squeezes a wide image onto the sensor, later unsqueezed to a widescreen ratio. Known for oval bokeh, horizontal flares and a distinctly cinematic look.

In a prompt: "anamorphic lens, horizontal lens flare, 2.39:1" signals the look even when the output ratio is 16:9.

See also: Aspect ratio, Lens flare

A lens that focuses very close, filling the frame with tiny subjects: insects, droplets, textures.

How much of the image, front to back, is in acceptable focus. Shallow depth of field isolates a subject; deep focus keeps everything sharp.

See also: Shallow depth of field, Deep focus, Aperture

Only a thin slice of the scene is sharp and the rest falls into blur, drawing the eye to the subject.

See also: Bokeh

Foreground, subject and background are all sharp at once, letting the audience choose where to look.

The opening inside a lens, written as an f-number. A low f-number (f/1.4) lets in more light and gives shallower depth of field.

The quality of out-of-focus areas, especially how points of light render. Anamorphic lenses make oval bokeh; spherical lenses make round.

Streaks or rings of light caused by a bright source shining into the lens. Used deliberately for warmth or a filmic feel.

The classic setup of a key light, a fill light and a back light. Nearly every other setup is a variation on it.

See also: Key light, Fill light, Rim light

The main light on the subject. Its direction and hardness decide most of the look of the shot.

A softer light from the other side that lifts the shadows the key light leaves. Less fill means more contrast.

A light from behind the subject that outlines their edges and separates them from the background.

Bright, even light with few shadows. Common in comedy, commercials and anything meant to feel open and safe.

Mostly shadow with pools of hard light and high contrast. The language of noir, horror and thrillers.

See also: Chiaroscuro

Strong contrast between light and dark that models form, borrowed from Renaissance painting.

A key light high and to one side that leaves a small triangle of light on the shadowed cheek. Moody and classical.

A light source visible in the frame, such as a lamp, candle or neon sign, that also lights the scene.

In a prompt: Naming practicals ("lit only by a desk lamp and a flickering TV") gives a model a physical reason for the light.

Light that appears to come from a believable source in the scene, like a window or streetlamp, even when it is really a film light.

The time shortly after sunrise and before sunset when sunlight is low, warm and soft.

See also: Blue hour

The twilight just before sunrise or after sunset, when the sky turns deep blue and city lights start to glow.

Light from a small or distant source that casts crisp, sharp-edged shadows, like midday sun.

See also: Soft light

Light from a large or diffused source that wraps around the subject with gentle shadow edges, like an overcast sky.

The subject is dark against a bright background, reduced to shape. It hides identity and emphasises outline.

Visible shafts of light made by haze, smoke or dust in the air. Models render it readily when you name the haze as well as the light.

How warm or cool light is, measured in kelvin. Tungsten bulbs are warm (about 3200K); daylight is cooler (about 5600K).

Divide the frame into a three-by-three grid and place key elements on the lines or where they cross, rather than dead centre.

The frame is balanced around a centre line. Formal and deliberate, it draws attention to anything that breaks the symmetry.

Lines in the scene, such as roads, rails or corridors, that guide the eye toward the subject.

The space between the top of the subject’s head and the top of the frame. Too much makes a subject look small; too little feels cramped.

See also: Lead room

Extra space in the direction a subject faces or moves. Without it, the frame feels like it is about to cut them off.

Empty area around the subject. It isolates, suggests loneliness, and leaves room for titles in a poster or thumbnail.

Using a doorway, window or out-of-focus object in the foreground to frame the subject and add depth.

The frame’s width compared to its height: 16:9 for most screens, 9:16 for vertical video, 2.39:1 for widescreen cinema, 1:1 for square.

In a prompt: Set the ratio in the model’s settings rather than only in the prompt. Models compose for the canvas they are given.

Where actors stand and move within a scene, planned against the camera. In AI video, blocking has to be written into the prompt or the reference frame.

Alternating between two characters facing each other, usually in over-the-shoulder or single shots. The backbone of dialogue scenes.

See also: 180-degree rule, Eyeline match

An imaginary line runs between two characters, and the camera stays on one side of it. Crossing it flips who faces left and right, and the audience loses their bearings.

A cut from a character looking at something to the thing they are looking at, with the gaze direction consistent across the cut.

A cut that joins two shots through a matching shape, movement or composition, linking them in meaning.

A cut between two similar shots of the same subject that makes time visibly skip. Used for energy or to compress time.

Sound leads or trails the picture across a cut. In a J-cut the next scene’s audio starts early; in an L-cut the current audio continues over the next shot.

A shot of something other than the main action, cut in and back out. It covers edits and adds context.

Keeping details consistent from shot to shot: costume, props, lighting, weather and a character’s face. The hardest thing to hold across AI-generated shots.

See also: Character reference, Character sheet

The set of shots filmed for one scene (wide, mediums, close-ups) so the editor has options. With AI you generate coverage deliberately, shot by shot.

Supplementary footage cut over the main action or narration: locations, details, atmosphere.

The story in one or two sentences: who wants what, what stands in the way, and why it matters.

A prose summary of the film, written before the script, that sets out story, tone and visual approach.

The written film: scene headings, action and dialogue in a standard format, roughly one page per minute of screen time.

A list of the story’s key moments in order. For a short film it is often the most useful bridge between idea and shot list.

Every shot in a scene with its size, angle, movement and action. In AI filmmaking the shot list becomes the list of prompts.

See also: Storyboard

A sequence of drawn or generated frames, one per shot, that shows what the film will look like before it is made.

See also: Shot list, Animatic

A storyboard cut together in time, often with temp sound, to test pacing before final footage exists.

A collection of reference images that pins down colour, light, texture and tone for the whole film.

One image that shows a character from several angles with their key features, used as the reference every later shot is checked against.

See also: Character reference, Continuity

The text instruction given to a model. For film work, a strong prompt reads like a line from a shot list: subject, action, setting, camera, light and style.

A list of things to avoid, supported by some models as a separate field. Models without one ignore negatives written into the main prompt, or even add them.

Generating a clip from a text prompt alone. Fast for exploring ideas, but the model decides everything you do not specify.

See also: Image-to-video

Generating a clip that starts from an image you supply. It fixes composition, character and lighting before any motion, which is why film workflows rely on it.

See also: First frame, Text-to-video

The image a video generation begins from. Everything visible in it is locked in at the start of the clip.

See also: Last frame, Image-to-video

An image the generated clip has to arrive at. With both a first and last frame, the model fills in the motion between two fixed points.

An image passed to a model alongside the prompt so it keeps a person’s face, build and costume consistent. Stronger than describing the character in words every time.

See also: Character sheet, Continuity

The number that initialises a generation’s random noise. Reusing a seed with the same prompt and settings can reproduce a result, on models that expose it.

Increasing an image or clip’s resolution after generation, adding detail rather than just enlarging pixels.

Matching a character’s mouth movement to dialogue audio, either generated with the clip or applied afterwards.