AI Image Aspect Ratios: Choose the Right Canvas Before You Generate

août 13, 2026

Aspect ratio is not an export setting you can safely postpone. It changes the image the model composes.

A wide 16:9 canvas encourages horizontal relationships, environment, and negative space beside a subject. A vertical 9:16 canvas encourages height, a centered figure, foreground-to-background depth, and room above or below. Cropping one into the other after generation often removes the visual reason the composition worked.

This guide explains the common ratios, how to prompt for them, and how to plan a small family of assets without stretching, padding, or repeatedly sacrificing the subject.

Aspect ratio versus resolution

Aspect ratio is the proportional relationship between width and height. Resolution is the number of pixels.

Both 1920×1080 and 1280×720 are 16:9. Both 1080×1350 and 800×1000 are 4:5. Increasing resolution adds pixel detail but does not change composition.

To calculate a ratio, divide width and height by their greatest common divisor. For practical work, you usually select a supported named ratio and choose an appropriate export resolution later.

Do not prompt “high resolution 9:16 landscape.” Landscape and 9:16 conflict. State ratio and orientation consistently:

Vertical 9:16 composition for a phone screen.

What each common ratio is good at

1:1 square

Square images are neutral and compact. They work for thumbnails, catalog tiles, profile-adjacent posts, album-like covers, and layouts where the card size is unknown.

Composition tendencies:

  • centered single subject;
  • balanced radial arrangement;
  • limited environmental context;
  • symmetrical or grid-based design.

Prompt example:

Square 1:1 editorial still life, one cobalt-blue kettle centered on a pale stone surface, soft shadow to the right, generous margin on all sides, no text.

Square is a useful master only when the subject can survive both horizontal and vertical crops. It is not automatically the most flexible choice.

4:5 portrait

Four-by-five offers more vertical room without the extreme height of phone-first 9:16. It suits portraits, product-plus-copy layouts, editorial covers, and feed images.

Composition tendencies:

  • upper-body or full-product framing;
  • room for a headline above or below;
  • visible foreground and background without excessive empty height.

Prompt example:

Portrait 4:5 fashion editorial, full jacket visible from collar to hem, subject slightly right of center, clean cream negative space in the upper left, eye-level camera.

If hands or product bases matter, explicitly ask that they remain fully inside frame. Portrait generators may crop at wrists, ankles, or table edges.

3:2 and 4:3 landscape

These classic photo shapes give more height than 16:9. They are useful for editorial photography, documentation, food, interiors, and scenes where the environment matters but should not become a cinematic strip.

Landscape 3:2 documentary photograph of a ceramicist at a workbench, both hands and the full wheel visible, shelves providing context behind, natural window light.

Choose 4:3 when you want a slightly more compact frame, especially for tablet or presentation contexts. Choose 3:2 for a familiar photographic feel.

16:9 landscape

Sixteen-by-nine is the standard widescreen shape for video and many hero sections. YouTube's official resolution and aspect ratio guidance identifies 16:9 as its standard desktop ratio and advises against baking padding or black bars into the file.

Composition tendencies:

  • subject placed on a third with context beside it;
  • two-subject relationships;
  • landscapes and interiors;
  • horizontal motion;
  • negative space for a page headline.

Wide 16:9 website hero, compact silver camera on the right third of a dark walnut desk, soft window light from left, uncluttered dark negative space across the left half, no text or logos.

For a web hero, check where the page will place text and buttons. Generate intentional negative space rather than covering important visual detail later.

9:16 vertical

Nine-by-sixteen fills a vertical phone viewport and suits short-form video covers, stories, and tall poster compositions.

Composition tendencies:

  • strong vertical lines;
  • full-body subjects;
  • depth stacking from foreground to background;
  • centered or lower-third focal points;
  • room for interface overlays.

Vertical 9:16 night street scene, cyclist centered in the lower half riding toward distant lights, tall buildings framing both sides, open dark sky in the upper quarter for interface-safe space.

Do not simply ask for “the same 16:9 scene but vertical.” Recompose. A side-by-side conversation may become an over-the-shoulder depth composition; a wide landscape may need one foreground subject rather than the entire horizon.

Choose the final placement before generating

Ask four questions:

  1. Where will the image appear?
  2. Will the container crop with object-fit: cover?
  3. Which overlays or interface controls occupy the edges?
  4. What must remain visible at the smallest breakpoint?

A website hero can be 16:9 on desktop but crop toward 4:5 on mobile. If one asset must cover both, keep the essential subject near the center and place expendable environment at the sides. If the design uses text on the left, the subject can sit right on desktop but may collide with text when centered on mobile.

The more different the target shapes, the stronger the case for separate compositions.

Prompt composition, not only the ratio

Adding “16:9” at the end is not enough. Describe how the ratio should be used:

Wide 16:9 composition. Place the person on the right third from waist up. Keep the entire left half quiet and low contrast for a headline. Preserve generous top and side margins.

For vertical:

Tall 9:16 composition. Show the full figure with shoes inside frame. Use foreground pavement, midground subject, and distant architecture to fill the height. Keep the top 15 percent simple for interface overlays.

Useful spatial phrases include:

  • centered with equal margins;
  • on the left or right third;
  • full object visible;
  • uncropped hands and feet;
  • negative space in the upper third;
  • horizon in the lower third;
  • straight-on label;
  • overhead flat lay filling the frame;
  • safe empty border around the subject.

These instructions turn ratio into layout.

Avoid stretching and baked padding

Never change a ratio by stretching width or height. Circles become ovals, faces widen, and product proportions become false.

Avoid adding black or colored bars simply to force dimensions unless the design intentionally includes a frame. Platforms and responsive web containers can handle native aspect ratios. YouTube explicitly advises against adding padding or black bars because its player adapts to the video's native shape.

The correct options are:

  • crop;
  • outpaint or extend;
  • recompose with a new generation;
  • place the native image inside a designed layout;
  • use a different asset for the target.

Each has a different cost.

Crop, outpaint, or regenerate?

Crop

Crop when the source has enough expendable space and the essential subject fits the target. Preview all target crops before approving the master.

Outpaint

Outpaint when you need more environment around a strong central subject. Expand one direction at a time and preserve the original pixels. Inspect repeating patterns, horizon continuity, shadows, and invented objects at the seam.

Regenerate or recompose

Regenerate when the target ratio needs a fundamentally different relationship. A wide product lineup may need a stacked vertical arrangement. A landscape with a tiny person may need a closer portrait framing.

Use the original as a visual reference when available, but write a new composition prompt for the target.

A multi-format production plan

For a campaign requiring 16:9, 4:5, and 9:16:

  1. Define the invariant: subject identity, wardrobe, product, palette, lighting, and location.
  2. Write a layout note for each ratio.
  3. Generate the most important placement first.
  4. Save the approved image as a reference.
  5. Generate or edit the secondary compositions using the invariant block.
  6. Add exact text as editable overlays when possible.
  7. Export at target pixel sizes without changing the native ratio.
  8. Test inside the real page or platform preview.

Example layout notes:

  • 16:9: subject on right, left negative space, waist-up.
  • 4:5: subject centered, knees-up, headline space above.
  • 9:16: full body lower center, tall architecture, safe top and bottom.

This creates a coherent family, not three identical crops.

Aspect ratio for text-heavy images

If text will be generated inside the visual, its container must be planned per ratio. A two-line headline in a wide left panel may need three lines at 4:5. At 9:16, copy may sit above the subject rather than beside it.

For exact or translated copy, keep the base image text-free and add editable typography. The AI image text rendering guide explains when integrated generated lettering is worth the trade-off.

Common mistakes

Selecting ratio after generation

The resulting crop removes hands, products, or environmental context. Decide placement first.

Asking for every target in one image

A prompt such as “works for square, portrait, and landscape” gives no single composition a clear priority. Generate a safe master or separate layouts.

Confusing platform dimensions with visual intent

Pixel requirements change; composition principles are steadier. Store the ratio and crop intent, then verify current platform export specifications at delivery time.

Placing important detail at edges

Responsive containers and platform UI can obscure edges. Keep critical faces, product marks, and readable text inside a safe central region.

Using aspect ratio to fix a weak prompt

A different canvas will not solve ambiguous subjects, conflicting styles, or too many actions. Ratio is one constraint, not the whole brief.

Pre-generation checklist

  • What is the exact destination and container behavior?
  • What ratio is native to that destination?
  • Where will copy and interface overlays sit?
  • Which subject details are non-negotiable?
  • Does the prompt describe subject placement and margins?
  • Will a crop preserve the idea, or is a separate composition needed?
  • Is important text editable?
  • Have all target placements been previewed before export?

Choose the canvas while the image is still an idea. That single decision makes prompting clearer, reduces destructive cropping, and produces assets that look composed for their destination instead of squeezed into it.

Lena Hoffmann

Lena Hoffmann