Composition and Framing Terms for Image Prompts

TL;DR: Composition and framing terms (wide shot, Dutch angle, negative space, rule of thirds) describe where the camera sits and how the frame is arranged, distinct from lens optics or lighting. Below are 26 real composition and framing terms, grouped by shot distance, camera angle, framing device, and compositional principle, each with a definition and a copy-paste fragment.

Two vendors put real weight behind part of this vocabulary. OpenAI's image-generation prompting guide instructs you to "Specify framing and viewpoint (close-up, wide, top-down), perspective/angle (eye-level, low-angle)" to control the shot (developers.openai.com/cookbook/examples/multimodal/image-gen-models-prompting-guide, accessed September 3, 2026). Google's Gemini API image-generation guide says to "Control the camera" using "photographic and cinematic language", naming "wide-angle shot, macro shot, low-angle perspective" as examples (ai.google.dev/gemini-api/docs/image-generation, accessed September 3, 2026). Shot distance and camera angle are real, vendor-named levers. The compositional principles further down this page — rule of thirds, leading lines, golden ratio — are craft convention passed down from photography and film, not terms either vendor names individually. Both are worth knowing; only one is documented.

One honesty note before the list: Prompt Architects writes the prompt, not the image. Nothing here renders a picture; these are words to paste into whichever tool does that.

What Do Composition and Framing Terms Actually Do in an Image Prompt?

They tell the model where to stand and how to arrange what's in front of it, independent of what lens it's "using" or how the scene is lit. "Nice composition" could mean anything; "medium shot, eye-level, subject in the left third of the frame" describes one arrangement. The two vendor quotes above are the closest thing to documented proof that this category of language does something: both companies name shot distance and viewpoint as levers a prompt writer controls, in guides aimed at people writing prompts for their own image models.

That said, composition is also where structured layouts get hardest. OpenAI's own guide lists this under its "Limitations" heading, not its best-practices section: "the model may have difficulty placing elements precisely in structured or layout-sensitive compositions" (platform.openai.com/docs/guides/image-generation, accessed September 3, 2026). Naming a shot type and an angle is a real lever. Expecting a model to hit an exact pixel-perfect layout from prose alone is asking for more precision than any current vendor promises.

How Do You Combine Composition Terms in One Prompt?

Layer them in order: shot distance, then camera angle, then one framing device or compositional principle. Two competing principles in the same sentence (symmetrical and rule-of-thirds, for instance) just make the model guess.

This is a companion piece to the anatomy of a great image prompt, which covers where composition sits inside a complete, six-part prompt. This page stays narrowly on the vocabulary itself.

Which Terms Describe Shot Distance?

Distance is how far the implied camera sits from the subject, which sets how much of the surrounding space makes it into frame.

Extreme wide shot (establishing shot)

The subject is small or absent; the frame is dominated by the environment, usually to set location and scale before anything else happens. extreme wide shot, tiny figure dwarfed by the landscape, environment fills the frame

Wide shot (long shot)

The full subject and a generous amount of surrounding space are both visible; the classic "here's the scene" framing. wide shot, full figure visible, generous surrounding environment in frame

Full shot

Tighter than a wide shot: the subject's entire body fills most of the frame, with only a little surrounding space, distinct from an establishing wide shot where the environment dominates. full shot, entire body visible top to bottom, minimal surrounding space

Medium shot

Roughly waist-up: enough to read body language and gesture without losing detail on the face, the default framing for a person talking or working. medium shot, waist-up framing, gesture and posture both readable

Medium close-up

Chest-up, tighter than a medium shot but still showing shoulders, the common framing for a portrait that isn't purely about the face. medium close-up, chest and shoulders visible, face the clear focal point

Close-up

Fills the frame with the face or a small object, cropping out almost everything else so expression or detail carries the shot alone. close-up, face fills the frame, background reduced to a soft edge

Extreme close-up

Tighter still: a single feature (eyes, hands, a small detail) fills the entire frame, past where a normal close-up would stop. extreme close-up, eyes fill the frame, every other detail cropped away

Which Terms Describe Camera Angle and Height?

Angle is where the implied camera sits vertically and how it's tilted, independent of distance; any angle can pair with any shot distance.

Eye-level shot

The camera sits at the subject's eye height, the neutral default that reads as neither dominant nor small. eye-level shot, camera at subject's eye height, neutral neither-up-nor-down framing

Low-angle shot

The camera looks up at the subject from below, which reads as powerful, imposing, or heroic by making the subject loom over the frame. low-angle shot, camera looking up from below, subject towering and dominant

High-angle shot

The camera looks down at the subject from above, which reads as small, vulnerable, or diminished. high-angle shot, camera looking down from above, subject appearing small and exposed

Bird's-eye view (top-down)

Directly overhead, near-vertical: flattens depth entirely and reads as a map or a diagram rather than a scene, distinct from a high-angle shot that still shows the subject's front. bird's-eye view, directly overhead camera, scene flattened into a map-like layout

Worm's-eye view

Directly from below, looking straight up: an extreme version of the low-angle shot, common for making architecture or trees loom overhead. worm's-eye view, camera on the ground looking straight up, structure towering overhead

Dutch angle (canted angle)

The camera itself is tilted so the horizon line isn't level, independent of height or distance; reads as tension, disorientation, or instability, and stacks on top of any other angle or shot distance. Dutch angle, tilted horizon line, framing reads unsettled and off-balance

Which Terms Describe Framing Devices?

A framing device uses an element inside the scene, not the camera's position, to direct attention toward the subject.

Frame within a frame

A rectangular or near-rectangular shape inside the scene (a doorway, a window, an archway) surrounds the subject, drawing the eye inward. subject framed by a doorway, natural rectangular border directing focus inward

Foreground framing

The broader technique behind a frame within a frame: any object placed between camera and subject that partially obscures the foreground to add depth and focus. foliage framing the foreground, subject sharp and centered beyond the blurred leaves

Negative space

Deliberately shaped and placed empty area that balances or isolates the subject, rather than empty space left over by accident. OpenAI's own prompting guide gives this as a worked layout instruction: "subject centered with negative space on left". subject positioned right of center, large negative space filling the left third of the frame

Fill the frame (tight crop)

The opposite instinct: the subject occupies almost the entire frame with little to no surrounding space, leaving no room to misread the composition. subject fills the frame edge to edge, no surrounding space, tight crop

Headroom (looking room)

The space left above a subject's head, or in the direction they're facing or moving; too little reads as cramped, too much reads as empty and adrift. comfortable headroom above the subject, extra space in the direction they're facing

Which Terms Describe Compositional Principles?

These are craft conventions borrowed from photography and painting: none of them are named individually in either vendor guide quoted above, so treat them as guidance you describe in your own words rather than a keyword a model is documented to parse.

Rule of thirds

Dividing the frame into a 3x3 grid and placing the subject along a gridline or at an intersection, rather than dead center, which most viewers read as more dynamic. subject positioned along the left third-line, not centered, dynamic off-balance composition

Golden ratio (Fibonacci spiral)

A proportion near 1.618:1, sometimes visualized as a spiral, used to place a subject slightly less symmetrically than the rule of thirds; more of a name-drop in casual prompting than a distinct instruction from the rule of thirds itself. golden ratio composition, subject placed along the spiral's natural focal point

Leading lines

Real or implied lines in the scene (a road, a railing, a shoreline) that guide the eye toward the subject rather than sitting as flat background detail. leading lines converging on the subject, a road stretching from foreground into the distance

Symmetry (centered composition)

The subject sits dead center with matching elements balanced on both sides, which reads as formal, stable, or deliberately artificial rather than dynamic. perfectly symmetrical composition, subject centered, matching elements mirrored left and right

Diagonal composition

The dominant lines or the subject's own axis run corner to corner rather than horizontally or vertically, which reads as more energetic or unstable than a level composition. diagonal composition, dominant line running from bottom-left to top-right, energetic unbalanced framing

Depth layering (foreground, midground, background)

Distinct visual information at three depths rather than a subject against a flat backdrop, which is what makes a static image read as spatially real. layered composition, sharp foreground detail, midground subject, soft distant background

Visual balance

Distributing visual weight (size, contrast, color) across the frame so no single area feels like it's dragging the eye away from the subject, achievable through symmetry or through counterweighting an off-center subject with a smaller contrasting element. visually balanced composition, off-center subject counterweighted by a smaller contrasting element on the opposite side

Vanishing point (one-point perspective)

Parallel lines in the scene appear to converge on a single point, usually centered, which reads as a strong, deliberate sense of depth: hallway and railway-track shots are the classic case. one-point perspective, parallel lines converging on a central vanishing point, strong depth

Wide Shot vs. Medium Shot vs. Close-Up: How Do They Actually Compare?

Three distances, three different jobs; mixing them up usually means the shot won't match the story you wanted to tell.

Which Composition Terms Get Confused, and What's the Real Difference?

Quick Reference: Which Term for Which Effect?

What Do Full Composition Prompts Look Like Combined?

Each stacks three to four terms in the order from the formula above.

Portrait, medium close-up, eye-level, subject positioned along the left
third-line, soft negative space filling the right side of the frame

Street scene, wide shot, high-angle, leading lines from the road converging
toward a small figure in the distance, layered foreground and background

Interior shot, subject framed by a doorway, symmetrical composition, one-point
perspective down a long hallway, vanishing point centered on the subject

Action scene, low-angle shot, Dutch angle, diagonal composition, subject
looming large against a tilted horizon

What Should You Do Next?

Pick a shot distance and a camera angle first — they're the two categories Google and OpenAI actually name in their own guides — then add one framing device or compositional principle, not three. For the vocabulary this page doesn't cover, lighting and camera and lens terms fill in the other two axes of a still-image prompt. For how all of it fits together in order, the anatomy of a great image prompt walks through a complete six-part prompt built one piece at a time.

Every vendor quote on this page was read from the vendor's own documentation on September 3, 2026.