Aspect-Ratio Control

Aspect-ratio control lets you specify the shape of a generated image — square (1:1), portrait (2:3), widescreen (16:9), or a custom ratio — rather than accepting a model's default. It matters more than it sounds: diffusion models are trained mostly on certain sizes, and asking for an unusual ratio affects not just framing but composition and quality. For SaaS builders, ratio control is the difference between a usable asset and a re-crop: social tools need vertical 9:16 for stories and 1:1 for feeds; ad platforms need multiple ratios of the same creative. Midjourney exposes it as --ar 16:9; API-based models take width and height in pixels, usually constrained to multiples of 64. Practical note: generating natively at the target ratio beats cropping a square, because the model composes for the frame — but pushing to extreme ratios like very wide panoramas can produce duplicated subjects, or 'twinning.' For those, generate closer to square and use outpainting to extend.