Midjourney is an amazing generative AI service, but to get the most out of it you need to know how to control it at a detailed level. With the right parameters you can determine the shape of the image, how much style the AI should add, what should be avoided, and how creative or strict it should be. Here you get a complete guide with all the parameters, clear explanations, examples, and information about which versions support which parameter.
Aspect ratio
Supported by Midjourney versions All
Controls the image’s proportions, that is the ratio between width and height. The default is square (1:1), but you can use for example 16:9 for wide images or 9:16 for mobile format. This is important if you are creating images for specific media or devices.
Parameter: –ar
Example: epic space battle --ar 16:9
Chaos
Supported by Midjourney versions: V5, V5.1, V5.2, Niji, V6, V7
Chaos determines how varied and unexpected the images become. A low value gives more consistent results, a high value gives more variation and stranger interpretations. Good if you want to experiment or let go of some control.
Parameter: --chaos
Min: --chaos 0
Max: --chaos 100
Example: a cat in a tuxedo riding a bike --chaos 95
Omni reference
Supported by Midjourney versions: V7
Use an image as a reference for how a character or an object should look in shape and silhouette. The AI tries to follow the shape, not the colors or style. This is useful if you want to reuse a character in several different environments.
Parameter: --oref [image-URL]
Example: portrait of a futuristic detective --oref [image-URL]
No
Supported by Midjourney versions: All
With --no you specify what you do not want included in the image. Want to create a forest without animals? Use --no animals. It is a simple way to have more control over the final result and avoid unwanted objects.
Parameter: --no
Example: fantasy landscape --no dragons
Personalization
Supported by Midjourney versions: V7
Use your saved style profile in Midjourney to influence the appearance of the image. The AI adapts based on what you have previously liked or used. Useful if you want to maintain a consistent style in your images.
Parameter: --p
Example: dreamlike jungle scene --p mysticMood01
Quality
Supported by Midjourney versions: V1 to V5.2, Niji
Not supported by: V6, V7
Determines how much computing power is used for the image. Lower values produce faster but simpler images, higher values produce more detail but cost more GPU time. Good when you want to make quick sketches or detailed final images.
Parameter: --q
Min: --q 0.25
Max: --q 2
Example: high-detail cityscape --q 2
Repeat
Supported by Midjourney versions: All
Used to run the same prompt multiple times automatically. It is an easy way to generate several variations of an idea without rewriting the prompt each time.
Parameter: --repeat
Min: --repeat 2
Max: --repeat 40
Example: concept art spaceship --repeat 5
Seed
Supported by Midjourney versions: All
A seed is a starting value for the AI’s random number generator. Same prompt + same seed = same image structure. Perfect if you want to create variations of a specific image or recreate something you made earlier.
Parameter: --seed
Min: --seed 0
Max: --seed 4294967295
Example: portrait of a sorcerer --seed 123456789
Stealth mode
Supported by Midjourney versions: All (requires Pro plan)
Prevents the image from being published on Midjourney’s website. Good for secret projects, corporate work, or when you simply don’t want to show the world what you are working on.
Parameter: --stealth
Example: product concept --stealth
Raw mode
Supported by Midjourney versions: V6, V7
Makes the AI follow the prompt more exactly, without automatically adding extra style or interpretation. Use it if you want maximum control and less “invention” from the AI.
Parameter: --raw
Example: technical diagram of a jet engine --raw
Stylize
Supported by Midjourney versions: V5, V5.1, V5.2, Niji, V6, V7
Determines how much artistic style the AI adds. Low values make the image more realistic or strict. High values make it more abstract, painted, or artistic.
Parameter: --s
Min: --s 0
Max: --s 1000
Example: dreamy fox in the woods --s 900
Style reference
Supported by Midjourney versions: V7
Use an image as a style template. The AI tries to match colors, brushstrokes, mood, and expression – but not the subject. Perfect if you want to create multiple images with the same feeling as an inspiration image.
Parameter: --sref [image-URL]
Example: magical illustration --sref [image-URL]
Tile
Supported by Midjourney versions: All
Creates images that can be repeated without visible edges. This makes the image seamless and perfect for patterns, backgrounds, or textiles.
Parameter: --tile
Example: looping geometric pattern --tile
Version
Supported by Midjourney versions: All
With --v you choose which model you want to use. Newer versions like V6 and V7 are better at details and understanding, but sometimes older versions provide a different style you prefer.
Parameter: --v
Example: epic dragon battle --v 5.2
Draft
Supported by Midjourney versions: V7
Makes a quick and cheap version of the image. It uses less GPU time and gives a “draft” rather than a finished artwork. Use when you want to quickly test ideas before putting more effort into the final result.
Parameter: --draft
Example: sketch of a fantasy throne room --draft
Weird
Supported by Midjourney versions: V5, V5.1, V5.2, Niji, V6, V7
The higher value you set for --weird, the stranger, more creative, and odd the images become. Low value = normal interpretations. High value = the AI invents really weird things. Perfect for those who like art that stands out.
Parameter: --weird
Min: --weird 0
Max: --weird 3000
Example: a sandwich with teeth on a chessboard --weird 2500
Fast mode
Supported by Midjourney versions: All
Fast mode means the image renders faster, but it costs more GPU time. Good when you are in a hurry, for example during live presentations or when testing many variations quickly.
Parameter: --fast
Example: branding mockup in brutalist style --fast
Image weight
Supported by Midjourney versions: All that support image prompt
Controls how much weight an uploaded image has compared to your text prompt. High value = AI follows the image more than the text. Low value = AI interprets the text more freely.
Parameter: --iw
Min: --iw 0.1
Max: --iw 2
Example: sci-fi knight based on sketch --iw 1.5
Relax mode
Supported by Midjourney versions: All
Relax means you let the AI work slowly, but without costing you GPU minutes (on certain subscription plans). Use when you have plenty of time and want to save on your renders.
Parameter: --relax
Example: detailed map of fantasy realm --relax
Turbo mode
Supported by Midjourney versions: All (requires Turbo plan)
Renders the image super fast. Only available for users with Turbo plan. Perfect for deadlines or if you work with AI images professionally at a high pace.
Parameter: --turbo
Example: ultra modern kitchen design --turbo
Niji
Supported by Midjourney versions: All (switches to the Niji model)
Niji is a specialized version of Midjourney trained to create anime, manga and East Asian styles. When you use --niji you switch to a completely different AI model, optimized for characters, colors and movement in Japanese style.
Parameter: --niji
Example: battle-ready anime girl --niji
Public mode
Supported by Midjourney versions: All
The opposite of --stealth. This makes your image publicly visible on Midjourney’s website. By default all images are public unless you specify otherwise.
Parameter: --public
Example: environment concept for sci-fi game --public
Motion
Supported by Midjourney versions: V7 (video beta)
If you work with video in Midjourney (beta feature) you can use --motion low or --motion high to control how much movement is added. Low = slower, more stable movements. High = more energy and pace.
Parameter: --motion
Example: abstract visual loop --motion high
Loop and end
Supported by Midjourney versions: V7 (video beta)--loop makes a video move in a perfect loop. --end lets you decide which frame to use as the ending. Perfect for music videos, digital art or background loops.
Parameter: --loop --end
Example: trippy forest animation --loop --end 120
Batch size
Supported by Midjourney versions: V7 (video beta)
Determines how many videos are generated from the same prompt. The higher the batch size the more variations. Good if you want to create many options quickly.
Parameter: --bs
Min: --bs 1
Max: Usually 6 (can vary)
Example: particle effects loop --bs 4
Style weight
Supported by Midjourney versions: V7
Determines how strongly your style reference (via --sref) will influence the result. Low value = the AI interprets more freely. High value = the image closely follows the style.
Parameter: --sw
Min: --sw 0
Max: --sw 1000
Example: portrait in cyberpunk style --sref [bild-URL] --sw 900
Style version
Supported by Midjourney versions: V7
Here you specify which version of the style reference system to use. This affects how the AI interprets your style image (if you use --sref). Useful if you want to experiment with how the style is transferred.
Parameter: --sv
Example: character card --sref [bild-URL] --sv 2
Video
Supported by Midjourney versions: V7 (video beta)
Used in Discord to generate video directly from a prompt. To use this, you need access to Midjourney’s video beta feature and be in the correct Discord server.
Parameter: --video
Example: dreamlike landscape --video
Sources
http://docs.midjourney.com/hc/en-us/articles/32859204029709-Parameter-List
