en Go beyond basic prompts. Master multi-prompting, weights, permutations, and --seed to get precise, reproducible, professional results every time.
Every great Midjourney prompt follows a structure. Understanding the parts lets you control each dimension of the output independently.
Full example prompt
a lone astronaut standing on a red desert planet, dramatic rim lighting, two moons visible, cinematic concept art --ar 16:9 --v 6.1 --style raw
a lone astronaut
The primary focus — what the image is about. Put this first.
standing on a red desert planet
Where the subject exists. Sets the scene and context.
dramatic rim lighting, two moons visible
Lighting is the single biggest quality lever in any image.
cinematic concept art
The visual treatment — photography, illustration, painting, etc.
--ar 16:9 --v 6.1 --style raw
Parameters that control format, model, and rendering behavior.
Midjourney weights earlier words more heavily. Always put your most important element first. "A golden retriever in a field" produces a better dog image than "a field with a golden retriever."
The double colon (::) separates a prompt into distinct concepts that Midjourney processes independently. This gives you precise control over how different elements are weighted and combined.
Without ::, Midjourney treats your entire prompt as one unified concept and blends all elements together.
With ::, each segment is processed as a separate concept. You can then assign weights to control how much each concept influences the final image.
A number immediately after :: sets the weight: ::2 gives that concept twice the influence of ::1.
a red apple on a wooden table
Midjourney treats this as one unified concept — you get an apple on a table.
a red apple :: on a wooden table
The double colon separates concepts. Midjourney now treats "a red apple" and "on a wooden table" as distinct ideas with equal weight.
a red apple ::2 on a wooden table ::1
The apple gets twice the weight of the table. The apple will be more prominent and dominant in the composition.
vibrant forest landscape ::2 autumn leaves ::-0.5
Negative weights subtract concepts. This generates a forest but actively suppresses autumn leaf imagery.
French
Italian
Spanish
German
Dutch
Arabic
Chinese (Simplified)
Japanese
Korean
Portuguese
Russian
Turkish
Swahili
Hausa
Yoruba
Polish
Sundanese
Kurdish (Kurmanji)Subject ::2 Style ::1Emphasize subject over style
Concept A ::1 Concept B ::1Equal blend of two ideas
Main ::3 Detail ::1 Background ::0.5Three-tier hierarchy
Desired ::2 Unwanted ::-1Suppress specific elements
Weights let you control exactly how much influence each concept has on the final image. Think of them as a mixing board for your visual ideas.
::0.5Very subtle influence — barely noticeable
Background texture, ambient mood
::1Default equal weight — balanced contribution
Standard multi-prompt baseline
::2Strong emphasis — this concept dominates
Primary subject, key visual element
::3Very strong — nearly overrides other concepts
When one element must be unmistakable
::-0.5Negative — actively suppresses this concept
Remove unwanted elements from output
cyberpunk city ::1 cherry blossoms ::1
Equal blend — both elements share the scene equally
cyberpunk city ::3 cherry blossoms ::1
City dominates — cherry blossoms become a subtle accent
cyberpunk city ::1 cherry blossoms ::3
Blossoms dominate — the city becomes a moody backdrop
Weights are relative to each other, not absolute. ::1 ::1 and ::2 ::2 produce identical results. What matters is the ratio between weights, not their absolute values.
Curly braces {} let you generate multiple prompt variations in a single command. Instead of running the same prompt 5 times with small changes, you define all the variations at once.
your prompt with {option1, option2, option3} in itMidjourney expands this into separate jobs — one for each option in the braces. You can use multiple brace groups in one prompt for multi-variable permutations.
Color variants — generate all colorways in one command
a {red, blue, green} sports car on a mountain roadExpands to:
a red sports car on a mountain roada blue sports car on a mountain roada green sports car on a mountain roadCharacter class exploration — test multiple archetypes at once
a portrait of a {warrior, wizard, rogue} in a fantasy settingExpands to:
a portrait of a warrior in a fantasy settinga portrait of a wizard in a fantasy settinga portrait of a rogue in a fantasy settingMulti-variable batch — style AND aspect ratio combinations
product photography, {minimalist, editorial, lifestyle} style --ar {1:1, 4:5, 16:9}Expands to:
product photography, minimalist style --ar 1:1product photography, editorial style --ar 4:5product photography, lifestyle style --ar 16:9Each permutation expansion counts as a separate job and uses GPU minutes. A prompt with 3 options generates 3 separate 4-image grids. On the Basic plan, use permutations sparingly. Standard and Pro plans handle batch generation much more comfortably.
Every Midjourney generation uses a random seed number to initialize the image. By specifying --seed, you can reproduce the same starting point — making results consistent and comparable.
A seed is a number (0–4294967295) that initializes the noise field Midjourney uses to generate images. The same seed + same prompt + same model version = the same image every time. Change any one of those three variables and the output changes.
Same seed + same prompt
Identical image
Same seed + different prompt
Similar composition, different content
Different seed + same prompt
Completely different image
You generated a perfect image but forgot to save the prompt
a misty mountain lake at dawn, watercolor style --seed 3847291 --v 6.1
Testing how a single prompt change affects the output
Compare: "golden hour lighting --seed 9182736" vs "blue hour lighting --seed 9182736"
Maintaining a character across scenes when you don't have a reference image
a young female astronaut, determined expression --seed 5647382 --v 6.1
React to any generated image with the ✉️ envelope emoji. Midjourney will DM you the job details including the seed number.
Click the three-dot menu on any generated image and select "Copy Job ID". Use /show [job ID] in Discord to retrieve the seed.
You can specify any number 0–4294967295 as your seed. Use a memorable number or generate a random one — the specific value doesn't matter, only consistency does.
Beyond the basics, these parameters give you fine-grained control over Midjourney's generation behavior, aesthetic interpretation, and output characteristics.
--chaosControls how varied the 4-grid results are from each other
Use 30–50 for early exploration, 0–10 for controlled refinement
--stylizeHow strongly Midjourney applies its aesthetic training
Low (0–50) = literal/raw. High (500–1000) = highly artistic/opinionated
--weirdIntroduces unusual, experimental aesthetics
Use 100–500 for creative exploration, higher values get very abstract
--tileGenerates seamlessly tiling patterns
Essential for fabric, wallpaper, and texture design
--stopStops generation at a percentage of completion
Lower values (50–70) create blurry, dreamlike, impressionistic results
--iwImage weight — how strongly an image prompt influences the output
Use with image URLs as prompts. Higher = closer to the reference image
Copy this template and fill in your values
[subject] [environment] [lighting] [art style] --ar [ratio] --v 6.1 --style raw --stylize [0-1000] --chaos [0-100] --seed [number]
Photorealism
--style raw --stylize 50 --chaos 0
Creative Exploration
--stylize 750 --chaos 40 --weird 200
Consistent Series
--seed 12345 --stylize 200 --chaos 0
Batch Testing
{option1, option2, option3} --chaos 10