me20
All news
/
News.Issue 9 ·

Midjourney image prompt vs style reference: which to use

An image prompt guides what is in the picture, a style reference only its look, and the Edit model keeps a subject. Which Midjourney reference to use.

By Olena, me20 Studio · 3 min read

Folk-art painting of rows of colourful fish with stripes, dots and scale patterns in red, orange, green and pink on a bright blue background
naive folk art acrylic painting, rows of whimsical colourful fish with stripes, dots and scale patterns in red, orange, yellow, pink and green, big round eyes, flat bright turquoise blue background, visible canvas texture, cheerful handmade mood --ar 16:9

In Midjourney, an image prompt borrows what a reference shows — its subject, composition and colours — while a style reference borrows only how it looks: colours, medium, textures and light, without copying objects or people. To keep the same product or character, use the Edit model. Pick the reference by its job.

What does an image prompt do?

An image prompt uses a picture as a source of inspiration for the content of a new image, not as something to copy.

  • Pair one image with text that describes the whole result, or blend several images, with or without text.
  • Image weight (--iw) sets how strongly the picture steers the result; the default is 1, and a higher value gives it more influence.
  • Midjourney's own tips: crop the reference to the aspect ratio of the final image, and describe in the text everything you want — not instructions for changing the reference.

What does a style reference copy?

A style reference transfers the overall look of an image — colours, medium, textures and lighting — and, in Midjourney's words, "doesn't copy objects or people".

  • It always needs a text prompt that says what to show.
  • Style weight (--sw) runs from 0 to 1000, with 100 as the default.
  • Instead of an image you can use a numeric style code (--sref plus the code), and several references can be mixed and weighted.

Need the same product or person? Use the Edit model

The Edit model generates from up to four reference images and replaces the older Omni Reference and Character Reference features. It works with Midjourney versions 8.1 and 8.2.

  • It tries to match the aspect ratio of the first image you upload; add --ar to change it.
  • If several characters get mixed up, the docs suggest combining them into one reference image.
  • Adding --raw reduces Midjourney's extra creative flair, so your text has more say.

Cheat sheet: which reference for which job

  • A layout, scene or colour mood → image prompt (--iw, default 1).
  • A look: palette, medium, light → style reference (--sw 0–1000, default 100).
  • The same subject or character → Edit model (up to 4 reference images).

Combine them, but give each one a job

Style references can be combined with image prompts and Edit model references in one prompt. Decide first what each picture is for — subject, layout or look — and say the rest in words. Midjourney also suggests describing the style in your text when you use a style reference with the Edit model.

New to the terms? Start with our AI image glossary and our guide to briefing an AI product photoshoot; our imagery portfolio shows the results. Want visuals in one consistent style? Tell us about your project.

P.S. The caption under the cover image is a ready-to-use prompt. Copy it into Midjourney, Higgsfield or any image generator and make your own version.

Also in News