Aspect ratio is not a crop applied after generation — it changes what the model composes in the first place. The same character and scene at 3:4 gives you a portrait framing; at 16:9 the model pulls back and builds an environment around the subject; at 1:1 it centers and tightens. Choosing the ratio is choosing the shot. Here is what each ratio does compositionally, which one fits which use, and why character art defaults to 3:4.
Ratio is composition, not cropping
When a model generates at 16:9, it does not make a 3:4 image and pad the sides — it plans a wide scene from the start: the subject moves off-center, the environment earns real estate, the light gets room to travel. Generate the identical inputs at 3:4 and the model plans a portrait instead: subject dominant, background compressed to context.
That is why cropping a generated image to a new shape almost always disappoints — you are cutting a composition that was balanced for a different canvas. Need a different shape? Change the ratio and generate again. With batch generation the re-roll is cheap.
The five ratios and what they're for
| Ratio | Shape | The model composes | Best for |
|---|---|---|---|
| 3:4 | Tall portrait | Subject-dominant, classic figure framing | Character art — the default |
| 9:16 | Full vertical | Full-body vertical, top-to-bottom flow | Phone wallpapers, story formats |
| 1:1 | Square | Centered, tight, symmetrical | Avatars, profile pictures, grids |
| 4:3 | Gentle landscape | Subject plus some environment | Scene shots that stay subject-first |
| 16:9 | Wide | Environment-forward, cinematic staging | Desktop wallpapers, establishing shots |
Three of these do most of the work in practice:
3:4 — the character default. A standing or seated figure naturally fills a 3:4 frame, which is why portrait painting settled on roughly this shape centuries before AI. If you have no specific destination in mind, this is the right pick, and it is what Gensomnia generates when you leave the setting alone.
9:16 — the phone canvas. Wallpapers and vertical-first formats. The model stretches composition vertically — full-body poses work; tight face shots leave awkward empty space above and below.
16:9 — the scene canvas. The subject becomes part of a staged environment. This ratio rewards location-rich prompts: name the setting and lighting explicitly (see the prompt structure guide) so the extra width has something to show.
Ratio interacts with everything else
With pose: wide frames invite reclining and horizontal poses; tall frames suit standing and seated ones. Asking for a full-body standing pose at 16:9 forces the model to either shrink the subject or crop them — specify pose and ratio compatibly.
With style: cinematic 16:9 flatters realistic rendering, where environment and light do the storytelling; anime and comic styles read strongest in subject-dominant tall frames.
With batch size: a new ratio effectively resets your hit rate — the model is solving a new composition problem. Budget a batch of four when switching shapes for a scene you care about.
Where the setting lives
Aspect ratio is the first control in the generation settings panel — the round settings button next to Generate — with five options and 3:4 preselected. Like batch size and the negative prompt, it is part of the Pro toolkit; the panel is visible to everyone, so you can see the full set of controls before upgrading.

Compose for the destination
Wallpaper, avatar, or portrait — pick the canvas before you generate.
Open the generatorFrequently asked questions
What is the best aspect ratio for AI character art?
3:4 portrait. It matches how a figure naturally fills a frame, which is why it is the default in character-focused generators.
Can I just crop my image to a different shape instead?
You can, but results disappoint — the composition was balanced for the original canvas. Re-generating at the target ratio produces a composition designed for that shape.
What ratio should I use for a phone wallpaper?
9:16. Pair it with a full-body pose; tight portraits leave dead space on a vertical canvas.
Does aspect ratio change how the model interprets my prompt?
Yes. Wide canvases push the model toward environment and staging; tall ones toward subject-dominant framing. The same prompt composes differently per ratio.
Why did my wide generation shrink the character?
A 16:9 canvas gives the model room it must fill; without an environment to build, it de-emphasizes the subject. Add location and lighting details to wide-format prompts.
Is aspect ratio control free on Gensomnia?
It is part of the Pro settings panel together with batch size and negative prompts. Default 3:4 generation is available to everyone.