Midjourney

Sets the width-to-height ratio of the generated image. Defaults to a square 1:1 when omitted; accepts any whole-number ratio such as 16:9 or 2:3.

Controls how varied and unexpected the initial four-image grid is. Higher values produce more diverse, surprising compositions; the default is 0.

Keeps a consistent character across images from a reference image, with a character weight (--cw) controlling how much of the reference is carried over.

Selects which Midjourney model version renders the image. Newer versions generally improve coherence, prompt-following, and detail; the default is the current latest.

Tells Midjourney what to leave out of an image. A nudge away from listed elements rather than a hard guarantee they will be absent.

Switches to Midjourney’s Niji model, tuned for anime and illustration styles. An alternative model choice to the default --v line.

Controls how much rendering time and detail a generation gets — it changes detail and GPU cost, not the output resolution.

Generates seamlessly repeating, tileable images for textures, patterns, fabrics, and wallpapers. The single output is one tile of the repeat.

Sets the starting noise used for a generation. The same seed with the same prompt and parameters produces a similar starting point — the basis of controlled variation.

Applies the visual style of a reference image (by URL) or a saved style code to a new prompt, without copying the reference’s subject matter.

Controls how strongly Midjourney applies its own default aesthetic. Low values stay literal to the prompt; high values produce more artistic, less literal images.

Introduces offbeat, experimental aesthetics into an image. Pairs with stylize, defaults to 0, and ranges far higher than most parameters.