Midjourney turns a written description into images. You type a prompt into the Imagine bar on the Create page at midjourney.com, press Enter, and it makes a set of four images. You need a paid plan before you can make your first image, according to Midjourney's Getting Started Guide. (Midjourney Docs: Getting Started Guide (opens in a new tab))
The short version of this guide: write less than you think, choose exact words, say how many of something you want, describe what should be in the picture rather than what should not, and put settings at the very end. Everything below comes from Midjourney's own documentation, plus notes from people who test new versions. (Midjourney Docs: Prompt Basics (opens in a new tab); Midjourney Docs: Parameter List (opens in a new tab))
Start with a short prompt
Midjourney's Prompt Basics page is blunt about length. Short, simple prompts usually produce the best images. It suggests thinking of a prompt as a quick snapshot of the idea, and it warns that long lists and detailed instructions can confuse the process. A prompt can be as small as one word. With fewer details you get more variety; with more details you get more control. (Midjourney Docs: Prompt Basics (opens in a new tab))
The docs give an example. A long, chatty request for blooming California poppies in bright orange, drawn with colored pencils, becomes one plain line. (Midjourney Docs: Prompt Basics (opens in a new tab))
Prompt 1: The short version
Colored pencil illustration of bright orange California poppiesChoose exact words and numbers
Word choice matters. The docs recommend specific synonyms over generic ones, such as huge or enormous instead of big. Numbers matter even more. A plural like cats leaves the count up to Midjourney. Three cats, or a collective noun like a flock of birds, makes the result predictable. (Midjourney Docs: Prompt Basics (opens in a new tab))
Prompt 2: Say how many
Three cats asleep on a sunny windowsill, soft morning lightAdd only the details you care about
When certain parts of the image matter to you, name them. Midjourney's docs list seven areas to think about: subject (a person, animal, object or place), medium (photo, painting, illustration, sculpture, doodle), environment (indoors, underwater, in the city), lighting (soft, overcast, studio, neon), color (vibrant, muted, pastel, black and white), mood (calm, playful, gloomy, energetic) and composition (portrait, headshot, closeup, bird's-eye view). (Midjourney Docs: Prompt Basics (opens in a new tab))
Tip: You do not need all seven. Pick the two or three that decide whether the image works for you, and let Midjourney's default style handle the rest.
Prompt 3: A detailed photo
Closeup photo of [subject] [doing something] in [place], [lighting] light, [color] colors, [mood] moodPut parameters at the end
Parameters are settings you type after the description. Midjourney's parameter list says they go at the end of the prompt, with a space before the two dashes and no commas or periods inside them. The most useful ones for beginners are --ar to change the shape of the image, --s (stylize) to control how much of Midjourney's own artistic style is applied, --chaos to get more varied results, and --raw for more control over the output. (Midjourney Docs: Parameter List (opens in a new tab))
For the current V8.1 model, a third-party prompt guide from Evolink lists the defaults: square 1:1 images, stylize at 100 on a scale of 0 to 1000, and chaos at 0 on a scale of 0 to 100. Stylize sets how strongly Midjourney's default look is applied, so lower values give a plainer result. (Evolink: Midjourney V8.1 prompt guide (opens in a new tab); Midjourney Docs: Parameter List (opens in a new tab))
Prompt 4: A wide image with a bit more style
Watercolor painting of a lighthouse on a cliff at sunset --ar 16:9 --s 250Tip: Use --ar 4:5 or --ar 9:16 for images meant for a phone screen, and --ar 16:9 for slides and desktop wallpapers.
Say what you want, not what you do not want
This is the mistake most people make. If you write a party with no cake, the docs warn, you may still get a cake, because mentioning a thing can pull it into the image. Describe the scene you want instead. When you really need to remove something, use the --no parameter, which Midjourney's parameter list describes as the way to exclude elements from an image. (Midjourney Docs: Prompt Basics (opens in a new tab); Midjourney Docs: Parameter List (opens in a new tab))
Prompt 5: Remove an object
Minimal living room with a gray sofa and tall windows --no plantsPick the right model
Midjourney's Models page lists V6, V6.1, V7 and V8.1, plus Niji 7 for anime and illustration. You can choose one for a single image by adding --v and the number at the end, for example --v 8.1, or set a default in the settings panel. V8.1 renders standard jobs several times faster than earlier versions, holds on to small details and can make 2K images with --hd. It does not support some older features, including Character Reference and Omni Reference. (Midjourney Docs: Models (opens in a new tab); Midjourney Docs: Parameter List (opens in a new tab))
Testers report that newer models want clearer direction. Heather Cooper, who writes the Visually AI newsletter, found that V8 rewards more detail than V7 did, and she recommends switching to --raw for controlled or photorealistic results. Midjourney's docs say the same about Raw: it removes default styling for stronger prompt adherence. (Visually AI (Heather Cooper): Midjourney V8 is live on alpha (opens in a new tab); Midjourney Docs: Models (opens in a new tab))
1. Write the core line
Describe the subject and the medium in one short sentence, such as a photo of a red bicycle against a white wall.
2. Add two or three details
Choose the lighting, mood or composition that matters most to you. Stop there.
3. Add parameters
Set the shape with --ar, and add --raw if the result looks too stylized.
4. Change one thing at a time
If the image is not right, change a single word or setting and run it again, so you know what made the difference.
Prompt 6: Text inside the image
Vintage travel poster for a mountain town with the words "Visit Alpine Falls" --v 8.1Tip: For words inside an image, put the exact text in quotation marks. Cooper reports that V8 renders quoted text with higher accuracy.
Midjourney is good at mood, lighting and style, and a short prompt often gives a more striking result than a long one. It is weaker when you need exact control over many objects at once, and results still vary from run to run. Treat the first set of four images as a draft, keep the version you like, and refine from there.






