Midjourney Prompts That Work: Basics, Parameters, Styles

Learn how to write Midjourney prompts that get consistent results. Use better words, image prompts, and key parameters like --ar, --no, --iw, and --p.

Silvena Written by Silvena A. AI tools & automation
18 min read
Midjourney Prompts That Work: Basics, Parameters, Styles

Midjourney is fast, but your prompt is the steering wheel. A few words can shift lighting, mood, lens feel, and how clean the details land.

This guide is for Midjourney users who want repeatable results, not lucky rolls. You will build prompts you can reuse, tweak, and share with teammates.

If you change 5 things at once, you cannot learn what mattered. Treat prompts like a small experiment, one variable per reroll.

What you will get from this

  • A prompt structure you can reuse across styles
  • Simple ways to fix common failures, like weird hands or mushy textures
  • Controls for composition, lighting, and realism without bloating the prompt
  • Quick checks to keep your look consistent across a set

The prompt mindset

Think in layers: subject, scene, camera, light, materials, and constraints. Start with the core image, then add only what improves clarity.

When you get a good result, save the recipe. Keep a short library of prompts for your favorite looks, then swap the subject and setting.

A clean starter template

    [subject], [action or pose], in [environment], [time of day], [lighting], [camera/lens], [style], [materials], [composition], [constraints]
  

Prompt basics

A Midjourney prompt is the text you give the model to describe the image you want to generate. It can be a single word, an emoji, or a short phrase with a few key details.

In practice, shorter prompts often perform better because they leave room for Midjourney’s default style to resolve the gaps. Add details only when they change the outcome you care about.

Write the prompt you would want to see in your final caption. Avoid instructions like “please,” “make it,” or “do not.” Describe the scene.

Build a strong prompt in 6 steps

  1. Start with the subject, name the main thing you want to see
  2. Add the medium, like photo, illustration, watercolor, 3D render
  3. Place it in an environment, like studio, street, forest, underwater
  4. Set the lighting, like soft window light, neon glow, overcast sky
  5. Lock the mood and color, like calm, tense, pastel, monochrome
  6. Define the framing, like close-up, wide shot, top-down, portrait

Short prompt, clearer result

Long shopping lists can dilute your intent. Start tight, then iterate with small changes so you can see what moved the output.

    Show a field of wildflowers, very bright orange, lots of petals, make it hand-drawn, colored pencils, add extra detail, vibrant and vivid, beautiful and stunning
  
    Colored pencil illustration of vivid orange poppies, simple background, soft shading, close-up
  

Choose words that narrow the image

Word choice matters. Specific nouns and strong adjectives steer the model more reliably than broad terms.

  • Swap vague size words for a precise synonym, like enormous, tiny, towering
  • Use a specific number instead of a plural, like three cats instead of cats
  • Try a collective noun, like a flock, a crowd, a bouquet, a swarm
Prompt details that usually pay off
What to specifyHelpful phrasing
Subjectperson, character, product, animal, landmark, vehicle
Mediumphoto, oil painting, ink sketch, vector poster, clay sculpture
Environmentstudio backdrop, rainy street, desert at dusk, kitchen interior
Lightingsoftbox, golden hour, neon rim light, foggy overcast
Colormuted tones, high contrast, pastel palette, black and white
Moodplayful, uneasy, calm, energetic, cinematic
Compositionheadshot, wide shot, top-down, symmetrical, negative space

Short vs detailed prompts

Short prompts

Best when you want variety and you are exploring the concept.

  • More variation across results
  • Faster to iterate and compare
  • Less risk of conflicting instructions
  • Less control over small details
  • May drift from your exact mental image
  • Harder to match a specific brand look

Detailed prompts

Best when a specific element must be consistent across a set.

  • Better control over key elements
  • Easier to repeat a look across multiple images
  • More predictable framing and mood
  • More time to write and maintain
  • Can overconstrain and flatten creativity
  • Conflicts are easy to introduce by accident

Say what you want, not what you do not want

Negative phrasing can backfire. If you describe an object you do not want, the model may still include it because you mentioned it.

To exclude elements, use the --no parameter at the end of the prompt and list items separated by commas.

    gouache still life on linen cloth, soft studio lighting --no fruit, citrus, berries
  
  • On web, append --no to the end of your prompt in the Imagine field
  • In Discord, append --no to the end of your prompt the same way
  • Prefer naming what you want when exclusion wording could be misread

Midjourney treats each word inside --no as an individual term. Avoid phrases that could accidentally imply restricted content; specify the clothing or style you do want instead.

Advanced prompts

Once basic prompting feels natural, you can push Midjourney further by combining text with image inputs and adding parameters. This is how you get more consistent style, tighter composition, and repeatable results across a series.

Advanced prompting works best when you keep the text prompt focused on the final image. Do not write instructions about changing the reference image, describe what the result should be.

The three building blocks

What changes when you add images and parameters
BlockWhat it doesCommon mistake
Text promptDefines subject, medium, scene, lighting, mood, and framingToo many adjectives that fight each other
Image promptsInfluence content, colors, and composition using one or more reference imagesAssuming the result will copy the image exactly
ParametersControls how the image is generated, parameters go at the end of the promptStacking many parameters before you know what each one changes

Image prompts that guide, not copy

Image prompts let Midjourney use a reference image as inspiration. The goal is to steer the model toward your target look, not to reproduce the original image.

If you need precise edits on your own image, use Midjourney’s Editor instead of trying to force exact changes through prompting.

Single image plus text

Pick one reference image that matches the vibe, then write text that describes the final image you want.

  1. Choose an image close to your intended composition
  2. Write a short text prompt for the final subject and setting
  3. Add only the details that must not change
        minimalist product photo of a matte black smart mug on a stone countertop, soft window light, shallow depth of field
      

Multiple images for blending

Use two or more images to blend visual elements, like palette, silhouette, and texture. You can add text if you need extra control.

  1. Select two or more images with one clear theme each
  2. Decide what each image contributes, like color or pose
  3. Add a text prompt to lock the missing details
        editorial portrait, clean background, soft rim light, confident expression
      

How to use image prompts

On web, you add images from the Imagine bar. You can upload new images, select past uploads, and pin them so they stay in place across multiple runs.

In Discord, you typically upload an image to a channel, then use its image link at the start of your prompt, followed by your text description.

Image prompt quality checks

  • Use supported formats: .png, .gif, .webp, .jpg, .jpeg
  • Crop the reference image to match your target aspect ratio
  • Use at least one image plus a text prompt, or use multiple images without text
  • Keep the text prompt about the final image, not a list of edits

Pros

  • Faster consistency across a series
  • Better control over palette and composition
  • Less guessing when matching an existing style

Cons

  • Can overconstrain and reduce variety
  • Low-quality references can bake in problems
  • Easy to forget what the image is steering, so you lose intent

Control influence with image weight

If you want more or less influence from your image prompt, use the image weight parameter --iw. Higher values make the reference image matter more.

    cozy reading nook, warm lamp light, knitted textures, soft shadows --iw 2
  
Image weight defaults and ranges
ModelDefaultRange
Version 710 to 3
Version 610 to 3
Niji 610 to 3

Parameters: the finishing controls

Parameters are tools that customize how the image is generated. They go at the end of the prompt and can change things like aspect ratio and how strongly the model stylizes your idea.

  • Keep parameters at the end so you can scan them quickly
  • Add one parameter at a time until you understand its effect
  • If you are using only images and no text, avoid parameters that require text intent, like --stylize or --weird
    clean product photo of a stainless steel water bottle, studio lighting, subtle reflection --ar 4:5
  

Where to get more help

For community troubleshooting and prompt ideas, look for Prompt Craft spaces on Discord and on the web.

Using parameters like a pro

Parameters are short controls you add to the end of a prompt to shape how Midjourney generates an image. They can change format, randomness, model behavior, privacy, and how strongly references influence the result.

Parameters belong at the end, after the description. Most parameter problems come from spacing, order, or accidental punctuation.

Parameter syntax rules

  1. Write your prompt text first, describe what you want to see
  2. Add one space
  3. Add parameters using two dashes, then the parameter name
  4. Avoid punctuation inside parameters unless the parameter requires it, like ratios
  5. Keep prompt text before parameters, do not add new text after parameters

Formatting examples

    vibrant California poppies --ar 16:9
  
    vibrant California poppies--ar 2:3
  
    vibrant California poppies - - ar 2:3
  
    vibrant California poppies --ar 2:3,
  
    vibrant California --ar 2:3 poppies
  

Quick checks before you reroll

  • Are all parameters at the end, with a single space before the first one
  • Did you avoid commas, periods, and stray symbols in parameter text
  • Did you keep prompt text together, with no extra words after parameters
  • Are you changing only one thing per test, so you can learn what mattered

Full parameter list

Use this list as a menu. Start with aspect ratio, then add one more control only when you can explain why it belongs in the prompt.

Common Midjourney parameters and what they control
ParameterSyntaxWhat it changes
Aspect ratio--aspect or --arOutput shape, default starts square
Chaos--chaos or --cVariation level across results from the same prompt
Omni Reference--orefReference a person’s likeness or an object form, replaces older character reference in v7
No--noExcludes listed elements, use commas to separate items
Personalization--profile or --pApplies your personalization profiles and moodboards to steer style
Quality--quality or --qDetail level and processing time tradeoff
Repeat--repeat or --rGenerates multiple image sets from one prompt
Seed--seedReproducibility for testing and controlled experiments
Stealth mode--stealthKeeps creations private on the Midjourney website
Raw mode--rawMore literal interpretation and less built-in styling
Stylize--stylize or --sHow strongly Midjourney adds artistic flair
Style reference--srefMatches the look and feel of another image
Tile--tileSeamless repeating patterns
Version--version or --vSwitches model version for different output behavior
Draft--draftLower-cost draft images in v7 for faster iteration
Weird--weird or --wPushes results toward unconventional, quirky outputs
Fast mode--fastUses Fast GPU speed mode
Relax mode--relaxUses Relax GPU speed mode
Turbo mode--turboUses Turbo GPU speed mode
Public mode--publicMakes creations public on the Midjourney website
Niji--nijiAnime and Eastern aesthetics focused model
Image weight--iwStrength of image prompts in the final result
Motion low and high--motion low and --motion highMotion settings for video generations
Video looping and end frames--loop and --endLooping video generation and a custom end frame
Batch size--bsNumber of videos generated per video prompt
Style weight--swStrength of style references
Style reference versions--svSelects the style reference version used
Video--videoEnables video generation in Discord

Personalization profiles

Personalization is a style assistant that adapts to your taste based on images you rank or like. When enabled, Midjourney can bias outputs toward your preferred aesthetics.

Personalization works with Midjourney versions 6 and 7. You can also create styles from curated collections using moodboards.

Unlock your global profile

  1. Open the Personalize page and locate your Global profile for your default version
  2. Start ranking pairs, pick the image you prefer each time
  3. Keep going until the progress bar completes and the profile unlocks
  4. Rank more pairs for better tuning, skipping less usually helps

Use personalization on web and Discord

On web

  • Turn personalization on using the button near the Imagine bar
  • Select one or more default profiles from the dropdown
  • Add --p at the end of a prompt to apply your defaults
  • Use a specific profile by appending --p with its ID
        modern desk setup, clean product photo, soft daylight --p
      
        modern desk setup, clean product photo, soft daylight --p pID
      

In Discord

  • Append --p to apply your default profiles
  • Use a specific profile ID, or reuse a code from a prior prompt
  • If you get an error, you likely need more ranking data first
        editorial portrait, soft rim light, neutral background --p code
      

Manage profiles and codes

You can create additional profiles for different looks. Each profile has an ID, and as the profile evolves it can produce new codes that label different versions of your taste model.

  • Use the profile ID to always apply the latest version of that profile
  • Use older codes to reproduce a past look you liked
  • Rename profiles to match your use case, like clean product, gritty street, pastel UI
  • Deleting a profile removes it from your list, but existing codes can still work

Stylize with personalization

When personalization is active, --stylize controls how strongly your profile influences the image. Lower values reduce the personalization effect, higher values push it harder.

    minimal logo mark on textured paper, studio light --p --stylize 50
  
    minimal logo mark on textured paper, studio light --p --stylize 300
  

Common mistakes to avoid

  • Mixing too many parameters before you understand their effects
  • Using image-only prompts with parameters that expect text intent, like stylize or weird
  • Writing negative phrases instead of using --no cleanly at the end
  • Putting parameters in the middle of a prompt and then continuing the description

Art of prompting

Midjourney can do a lot with a short prompt, but the best results come from choosing the right concept words. Mediums, time periods, emotions, colors, and environments all steer the image in different ways.

The goal is not to stack buzzwords. The goal is to pick a few strong signals that describe the final image clearly.

When results look random, your prompt is usually too broad. Narrow one thing first, like medium or lighting, then reroll.

Turn one idea into 10 variations

  1. Write a stable subject you can reuse, like “sleeping cat on a windowsill”
  2. Choose one category to vary, like medium or time period
  3. Make a small set of options, five to 10 words max
  4. Run each option with the same subject text
  5. Save the top two, then vary a second category
    [subject], simple background, clear focal point, soft light
  

Artistic mediums that change everything

Medium words often have a bigger impact than extra adjectives. They push line quality, texture, color behavior, and how “finished” the image feels.

Medium keywords to try
Medium wordWhat it tends to addPrompt fragment
Block printBold shapes, carved texture, limited tonesblock print style
Ballpoint pen sketchHatching, fine lines, casual draft energyballpoint pen sketch
CyanotypeBlueprints, high contrast, photographic silhouettescyanotype print
GraffitiSpray texture, bold outlines, street vibegraffiti mural style
RisographInk grain, slight misregistration, poster feelrisograph print
Ukiyo-eWoodblock look, stylized linework, flat colorukiyo-e
WatercolorSoft edges, paper texture, flowing blendswatercolor illustration
Pixel artChunky pixels, retro game vibe, sharp silhouettespixel art
Cross stitchThread texture, grid structure, handmade craftcross stitch
Oil paintingBrush strokes, richer shadows, painterly depthoil painting

Pros of medium-first prompting

  • Faster style control than stacking adjectives
  • More consistent texture across a series
  • Clearer results with fewer words

Cons to watch for

  • Some mediums can overpower subject detail
  • Too many style words can clash and muddy output
  • Highly stylized looks may reduce realism

Time periods as a design shortcut

Adding an era nudges clothing, materials, color treatment, and graphic language. Use it when you want a cohesive mood without writing long descriptions.

Era cues you can drop into prompts
Time periodWhat it impliesPrompt fragment
1400sIlluminated manuscript feel, ornate detail1400s illustration
1700sClassic portrait cues, formal styling1700s portrait
1920sArt deco geometry, bold poster shapes1920s art deco
1950sMid-century polish, clean advertising look1950s magazine ad
1980sNeon palette, synth vibe, strong contrast1980s neon aesthetic
2000sEarly web polish, glossy highlights2000s product shot

Emotion, color, environment: the feel knobs

These words do not just change mood. They change lighting, pose, contrast, and composition decisions inside the model.

High-signal emotion words

  • Shy
  • Determined
  • Joyful
  • Angry
  • Sleepy
  • Sad

Color directions that steer style

  • Pastel palette
  • Duotone
  • CMYK print look
  • Sepia
  • Neon
  • Grayscale

Environment prompts to try

  • Tundra
  • Salt flat
  • Jungle
  • Desert
  • Forest
  • Cave
  • City
  • Ocean
    [emotion] [color direction] [medium] cat in a [environment], simple background, clear focal point
  

Image size and resolution

Image size is measured in pixels, width by height. When you upscale, Midjourney outputs a fixed pixel size for the selected model and upscaler.

For screens, the DPI number you see in software is not the quality. What matters is pixel dimensions and how large the image is displayed.

Printing: pixels to inches

For print, DPI matters because it controls how many dots land in each inch of paper. A common target for sharp prints is 300 DPI.

Example: a 2048 x 2048 px image can print at about 6.8 x 6.8 inches at 300 DPI.

Quick print sizing from pixel dimensions
PixelsDPIApprox. print size
2048 x 20483006.8 x 6.8 inches
2048 x 204815013.7 x 13.7 inches

Prepare an image for printing

  1. Open the upscaled image in an editor, like Photoshop
  2. Change the resolution to 300 DPI
  3. Turn resample off so pixel dimensions stay the same
  4. If you need larger prints, upscale to add pixels before increasing print size

Aspect ratio choices

Aspect ratio describes width compared to height, written as two numbers like 1:1 or 16:9. Midjourney starts square by default, and you can change it with --ar.

Aspect ratio is not the same as final pixel dimensions. Pixel size depends on the model and upscaler, while aspect ratio sets the shape.

Set aspect ratio with a parameter

    [your prompt] --ar 16:9
  
Common aspect ratios and where they fit best
Aspect ratioShapeTypical use
1:1SquareSocial posts, icons, centered compositions
4:3Light landscapeClassic framing, general illustration layouts
2:3PortraitPrint-friendly photos, posters, framed work
16:9WidescreenHeaders, video thumbnails, cinematic scenes
9:16VerticalMobile stories, vertical posters, reels-style compositions

Final notes

Great Midjourney prompts are not long, they are intentional. Pick a clear subject, add only the details that change the outcome, then iterate like a designer.

Save your winners as reusable recipes. Over time you will build a small prompt library for your favorite styles, formats, and use cases.

When you feel stuck, simplify. Remove half the words, keep the strongest two cues, and rerun. Clarity beats complexity.

FAQ

How long should a Midjourney prompt be?

Start short, one clear subject plus a few high-impact details like medium and lighting. Add more only when you can name the specific problem you are trying to fix.

Why do my prompts feel inconsistent across rerolls?

Your prompt may be too broad, or you may be changing multiple variables at once. Lock one element, like medium or aspect ratio, then adjust one detail per run so you can see what actually caused the shift.

When should I use image prompts instead of text-only prompts?

Use image prompts when you need stronger control over palette, composition, or a specific visual vibe. Pair one image with a clear text description of the final image, or use multiple images to blend visual elements.

How do I use parameters without breaking my prompt?

Put all parameters at the end, add one space before the first parameter, and avoid punctuation unless required for a value like a ratio. Keep prompt text before parameters and do not add new text after them.

How can I avoid unwanted objects showing up?

Describe what you do want in the scene first. If you still need exclusions, add a clean --no list at the end and separate items with commas, keeping the list specific and unambiguous.

Can I print Midjourney images, and what size will look sharp?

Yes. For sharp prints, a common target is 300 DPI. Print size depends on pixel dimensions, so if you want a larger print, you will usually need more pixels through upscaling before increasing physical size.

Style and keyword reference

If you want a fast way to explore visual directions, use this community reference of Midjourney styles and keywords with examples.

  • Use it to pick a medium, theme, palette, or design style keyword
  • Compare examples to learn what a term actually does visually
  • Turn the best keywords into a small prompt library for repeatable results
How to use a style reference without getting lost
What you needWhat to pull from the referenceHow to test it
A new look fastOne medium word plus one palette directionRun three variations, keep subject identical
Consistency for a setTwo to three keywords that repeat across imagesLock aspect ratio, change only the subject
Better compositionDesign style terms, framing cues, poster languageCompare wide vs portrait with the same keywords

Resource: Midjourney styles and keywords reference

Scroll to Top