✅ Verified up-to-date [Last edited: 05/21/2026]
<aside> <img src="/icons/magic-wand_gray.svg" alt="/icons/magic-wand_gray.svg" width="40px" />
When unwanted objects or details keep appearing in your Midjourney images, don’t start by telling Midjourney what to remove. Start by changing the prompt so it describes what the image should visibly contain.
This page teaches two main strategies:
Prompt what it looks like, not what it is.
Some words bring visual baggage with them: unicorns bring horns, cowboys bring hats, cafés bring coffee cups. If the word itself triggers the thing you do not want, avoid that word and describe the visible instead.
Use --no when positive prompting is not enough.
After you have built the strongest positive prompt you can, add the --no parameter at the end of the prompt to suppress unwanted words or phrases.
For MJ versions 5.x and 6.x, you can also use negative prompt weights in a multi-prompt, such as mushrooms::-0.5, to reduce the influence of an unwanted concept.
The basic rule is: describe the image you want first, then use --no or negative weights as cleanup tools.
</aside>
<aside> <img src="/icons/help-alternate_blue.svg" alt="/icons/help-alternate_blue.svg" width="40px" />
</aside>
Follow these two steps.
Certain words bring visual details with them. Unicorns have horns, cowboys have hats, people in cafes have coffee. So we need to avoid those words: unicorn, cowboy, cafe. To do that, we think about what the image looks like and describe it. We avoid saying what it actually is, because those words would trigger the undesirable details.
Midjourney doesn’t interpret our words the way people do, so it struggles with absence language like “without,” “no,” “not,” “hasn’t,” “doesn’t,” “isn’t,” “aren’t,” “lacks,” “missing,” “removed,” “instead of,” etc. We notice much higher rates of success when we avoid using absence language and use presence language instead. That’s the first step.
❌ WHAT IT IS…
A unicorn without a horn.
A pegasus with no wings.
A cowboy without a hat.
A woman waiting in a cafe hasn’t received her coffee yet.
✅ WHAT IT LOOKS LIKE…
A white horse with a flowing mane stands in a moonlit forest.
A white horse hovering in the sky.
A rugged man with tousled hair wearing a red kerchief.
A woman with brown eyes and hair in a messy bun is waiting at a table in front of a window looking out onto a springtime parking lot. On the empty wood table is a folded newspaper, her iPhone, and a vase of little yellow flowers. People sit at other tables nearby.
Certain words carry visual baggage. A unicorn brings a horn. A cowboy brings a hat. A café brings coffee cups, tables, pastries, and customers. When those attached details are unwanted, avoid the trigger word and describe the visible scene instead.
So instead of naming what it is, describe what the image looks like:
A white horse with a flowing mane stands in a moonlit forest.
instead of a unicorn without a horn
Avoid the archetype when the archetype contains the problem. Describe the visible replacement.
--no parameter.First, develop the prompt to its strongest version using the tips in step #1. But, when even positive prompting isn’t giving you what you want, add the --noparameter to that prompt.
At the end of the prompt, add --no word or --no phrase for the things you want to suppress.