Most Midjourney users write a prompt, hit enter, and accept whatever comes back - without ever touching Midjourney parameters.
The users getting consistently professional results do something different. They use parameters.
Parameters are modifiers you add to the end of any Midjourney prompt. They control aspect ratio, artistic interpretation, randomness, model version, image references, character consistency, and dozens of other variables that determine the final output.
Learning parameters is the single biggest quality jump available to intermediate Midjourney users. This guide covers every parameter worth knowing - what it does, when to use it, and real examples of how it changes your results.
Learning parameters is the single biggest quality jump available to intermediate Midjourney users.
How parameters work
Parameters always go at the end of your prompt after the descriptive text. They start with two dashes and a keyword followed by a value.
Basic structure:
[your prompt text] --parameter value --parameter value
Example
'Portrait of a woman in a white linen shirt, soft window light, minimal --ar 4:5 --v 8.1 --stylize 200'
The essential parameters - use these on every prompt
--ar (aspect ratio)
What it does: sets the dimensions of the output image. This is the most used parameter and one of the most important for professional work.
How to use it: --ar width:height
Standard ratios and when to use them:
--ar 1:1 - square. Instagram posts, profile images, Amazon product listings.--ar 4:5 - portrait rectangle. Instagram feed posts, the most common social media format.--ar 9:16 - tall vertical. Instagram Stories, TikTok, Pinterest.--ar 16:9 - wide landscape. YouTube thumbnails, website banners, presentations.--ar 2:3 - portrait. Pinterest, print, editorial.--ar 3:2 - landscape. Standard photography ratio, website imagery.--ar 21:9 - ultra wide. Cinematic, website hero banners.
Pro tip: always set --ar before you generate. Cropping after the fact almost always loses important composition elements that Midjourney placed intentionally.
Cropping after the fact almost always loses important composition elements that Midjourney placed intentionally.
Custom ratios work too. --ar 5:7 or --ar 7:4 are both valid. Any whole number ratio is accepted.
--v (version)
What it does: specifies which Midjourney model version generates your image. Different versions have different aesthetic qualities and capabilities.
Current versions:
- --v 8.1: the latest version as of 2026. Strongest overall quality, most photographic, best prompt following, sharpest detail, most accurate material rendering.
- --v 8: strong quality, slightly different aesthetic from 8.1. Some users prefer it for certain artistic styles. Good alternative when 8.1 feels too clinical.
- --v 7: previous generation. Still produces excellent results. Some users prefer its aesthetic for fashion editorial and artistic content.
- --v 6.1: older model. Painterly quality works well for illustration-adjacent content. Rarely needed for professional brand photography.
- --niji 6: specialized model for anime and illustration styles. Not useful for photographic brand content.
When to use each version:
- --v 8.1: default choice for all professional brand photography, product shots, and lifestyle imagery. Use this unless you have a specific reason not to.
- --v 8: when you want slightly more artistic rendering than 8.1 without going back to V7.
- --v 7: when you specifically prefer V7's aesthetic or are following a prompt that was built and tested on V7.
- --v 6.1: niche use cases only, artistic and painterly content.
--stylize (creative interpretation)
What it does: controls how much creative interpretation Midjourney applies to your prompt. The most important parameter for controlling the balance between prompt accuracy and aesthetic quality.
Range: 0 to 1000. Default: 100.
How each range behaves:
0 to 50: extremely literal. Follows your prompt as precisely as possible. Minimal aesthetic enhancement. Results can look flat or clinical but are highly accurate to your description.
51 to 200: balanced. Follows your prompt accurately while adding professional aesthetic quality. This range produces the most consistently useful results for brand photography. The default of 100 sits in this range for good reason.
201 to 400: elevated. Adds noticeable artistic quality, richer lighting, more sophisticated color relationships. Prompt is still followed accurately but Midjourney is making more creative decisions.
401 to 700: artistic. Strong aesthetic interpretation. Images look beautiful but begin to drift from your exact prompt description. Good for editorial and artistic content where the prompt is a direction not a specification.
701 to 1000: maximum interpretation. Midjourney applies its strongest aesthetic preferences. Your prompt is treated as loose inspiration. Results are often stunning but rarely match your precise intent.
Recommended starting points by use case:
- Product photography:
--stylize 100 to 150
- Lifestyle brand imagery:
--stylize 150 to 250
- Fashion editorial:
--stylize 250 to 400
- Artistic and abstract content:
--stylize 400 to 600
--chaos (variation)
What it does: controls how different the 4 images in each generation grid are from each other. Low chaos produces 4 similar variations. High chaos produces 4 dramatically different interpretations of the same prompt.
Range: 0 to 100. Default: 0.
When to use low chaos (0 to 20): when you have a working prompt and want consistent reliable variations of the same concept. Brand photography, product shots, any professional work where consistency matters.
When to use high chaos (50 to 100): when you are exploring and want to see many different interpretations quickly. Good for early creative exploration, finding unexpected directions, breaking out of predictable results.
Pro tip: run high chaos first to find a direction you love, then drop to low chaos to refine that specific direction with consistent variations.
--raw
What it does: disables Midjourney's built-in aesthetic filter. Without --raw Midjourney applies its own sense of beauty and polish to every output - smoothing skin, enhancing colors, adding a characteristic sheen. With --raw it follows your prompt more literally without this automatic enhancement.
When to use it:
- When you want genuinely photographic results without the Midjourney aesthetic overlay
- When skin is looking too smooth or artificial
- When colors are being enhanced beyond what you want
- When you want more realistic imperfections and natural texture
- When combined with film stock language for authentic analog photography feel
When not to use it: when you want Midjourney's aesthetic quality applied automatically. For artistic and editorial content --raw often produces less visually impressive results.
Example: adding --raw to a portrait prompt combined with 'visible pores, asymmetrical features, natural skin texture, fine hair catching backlight' produces results that are significantly more photographic and less AI-looking.
--seed
What it does: sets a specific random seed number that makes generations reproducible. When you use the same prompt with the same seed you get the same image every time.
How to find a seed: after generating an image you like, react to the message with an envelope emoji in Discord. Midjourney sends you the seed number for that generation in a direct message.
When to use it:
- When you want to reproduce a specific result
- When iterating on a prompt and want to isolate the effect of prompt changes from randomness
- When sharing prompts with others and want them to reproduce your exact result
Range: any whole number. Common practice is to note seeds for your best results and reuse them when you want similar outputs.
--weird
What it does: introduces unusual and unexpected elements into the generation. Higher values produce increasingly strange and surreal results that deviate from conventional aesthetics.
Range: 0 to 3000. Default: 0.
When to use it: creative exploration, artistic content, breaking out of expected results, generating unusual brand visuals that stand out. Not useful for commercial product photography or conventional lifestyle imagery.
Start low: --weird 100 to 300 adds interesting unexpected details without making images incomprehensible. Values above 1000 produce genuinely surreal results.
--draft
What it does: generates images at lower quality and cost for testing purposes. Draft mode uses significantly fewer GPU credits - approximately one fifth of a normal generation.
When to use it: testing new prompts before committing to full quality generations. If you are iterating on a prompt concept run 5 to 10 draft generations to find the right direction then switch to full quality for the final versions.
This is the most important parameter for managing your Midjourney credit usage efficiently. Never run full quality generations on untested prompts.
Never run full quality generations on untested prompts.
Reference parameters - the advanced tier
--sref (style reference)
What it does: transfers the aesthetic style of a reference image to your generation. Midjourney analyzes the visual style of the reference - its color palette, lighting quality, texture, mood - and applies those qualities to your prompt.
How to use it: --sref [image URL]
What transfers: color palette and temperature, lighting quality and mood, overall aesthetic and texture, compositional tendencies.
What does not transfer: specific subjects, people, products, or objects from the reference image. --sref transfers style not content.
When to use it: when you have a reference image that captures the aesthetic you want and find it difficult to describe in words. Brand moodboard images, competitor references, magazine tear sheets.
Strength control: --sref [URL] --sw 100 sets the style weight. 0 to 1000, default 100. Higher values apply the reference style more strongly.
--oref (omni reference)
What it does: maintains the visual identity of a specific subject - a person, product, or object - across multiple generations. Available in V7 only.
How to use it: --oref [image URL]
What it does well: keeping a specific product recognizable across different scenes and lighting setups. Maintaining a character's appearance across multiple images. Building a consistent visual library featuring the same subject.
What it struggles with: perfect reproduction of fine details like text and labels. Complex multi-element products. Very small or intricate design elements.
When to use it: product photography where you need the same product in multiple scenes, character-consistent content, any work requiring visual continuity across multiple images.
Strength control: --oref [URL] --ow 100. 0 to 1000, default 100. Higher values maintain the reference more strongly but can produce stiffer less natural results.
The complete parameters reference
Build this into a quick reference you can paste at the end of any prompt:
- Aspect ratio:
--ar 1:1 / 4:5 / 9:16 / 16:9 / 2:3 / 3:2
- Version:
--v 8.1 (latest) / --v 8 / --v 7 / --v 6.1 / --niji 6
- Stylize:
--stylize 0-1000 (default 100)
- Chaos:
--chaos 0-100 (default 0)
- Weird:
--weird 0-3000 (default 0)
- Raw mode:
--raw
- Draft mode:
--draft
- Seed:
--seed [number]
- Style reference:
--sref [URL] --sw [0-1000]
- Omni reference:
--oref [URL] --ow [0-1000]
Parameter combinations that work
For photorealistic product photography:
Build prompts with parameters automatically
Our free Midjourney Prompt Builder includes a dedicated parameters section with visual selectors for aspect ratio, version, stylize slider, chaos slider, and toggles for raw and draft mode. Every parameter you set updates the live prompt preview instantly. When you are done hit Enhance with AI to get 3 variations - subtle, standard, and cinematic - each with parameters tailored to that style.
CropUp Ai