Midjourney v6 Parameters Guide: --ar, --style, --chaos Explained

Midjourney v6 Parameters Guide: --ar, --style, --chaos Explained

Here's an honest admission: for the first four months I used Midjourney, I barely touched the parameters. My prompts ended without any -- flags at all. And you know what? I got decent images. Midjourney's defaults are good.

But once I understood what each parameter actually did — not just the description from the docs, but the felt difference in practice — my output quality jumped in a way that better prompt-writing alone never achieved. The parameters aren't just settings. They're a second creative vocabulary on top of your words.

This guide covers every significant v6 parameter with examples, comparison tables, and my honest opinion about when each one is actually worth using.

How Parameters Work in Midjourney

Parameters are flags you add to the end of your prompt, after your text description. They always start with -- and are separated by spaces.

a golden field at sunset, painterly --ar 16:9 --stylize 300 --style raw

Order doesn't matter among parameters. What matters is that they come after your prompt text, not within it.

--ar (Aspect Ratio)

This is the parameter you'll use on almost every single image. It sets the width-to-height ratio of your output.

serene lake surrounded by pine forest --ar 16:9

fashion portrait, editorial magazine --ar 2:3

mobile phone wallpaper, vibrant abstract --ar 9:16

Aspect Ratio Reference Table

Critical point: Aspect ratio changes composition, not just cropping. Ask Midjourney for the same portrait at 1:1 and 2:3 and you'll get genuinely different images, not the same image resized. At 2:3, it knows to include more of the figure. At 1:1, it'll probably emphasize the face. This is worth thinking about before you generate.

--style (Style Mode)

The --style parameter in Midjourney v6 has two main values that matter: the default (no flag needed) and raw.

--style raw

This is the one I reach for most. It turns off Midjourney's built-in aesthetic processing, which normally softens images, adds a slight painted quality, and generally tries to make things look "beautiful" according to the model's training.

weathered barn in a field, overcast sky, photography --style raw

close-up of old hands holding soil, natural light --style raw

Without --style raw, these prompts would likely produce images that look slightly over-polished — the barn too perfectly lit, the hands too clean. With it, you get something rawer and more photographic.

When to use --style raw:

• Photorealistic photography prompts
• Documentary and editorial images
• When your previous prompt came out looking too "AI-generated pretty"
• Product photography that needs to look real

When to skip it:

• Fantasy and concept art
• Stylized illustration
• Anything where you want Midjourney's aesthetic interpretation

Stylized Modes with --style

Beyond raw, you can append style values from certain aesthetic presets when prompted correctly:

portrait of a musician, --style scenic

--style scenic was available in v5 and pushes toward cinematic, atmospheric landscapes and scenes. Its behavior in v6 is more nuanced — test it if you want something more cinematic than standard.

--stylize (or --s)

This is one of the most misunderstood parameters. --stylize controls how much of Midjourney's aesthetic judgment gets applied to your image. Not how stylized the art looks — how much the model overrides your literal description to produce something "better" by its own standards.

Range: 0 to 1000. Default: 100.

cozy reading nook by a window, afternoon light --s 50

cozy reading nook by a window, afternoon light --s 750

The first prompt will stay closer to a literal interpretation — you'll see a room, a window, some books. The second will produce something more visually dramatic — possibly gorgeous, but likely with more invented elements, more dramatic lighting, more composition choices that weren't in your prompt.

--stylize Value Guide

My honest take: I rarely go above 400. Past that, the images often look spectacular but they stop being what I asked for. The sweet spot for most use cases is 150–300 — you get Midjourney's genuine aesthetic intelligence without losing control of your prompt.

--chaos (or --c)

--chaos controls variation between the four images in your initial generation grid. Low chaos means all four images are similar compositions of your prompt. High chaos means each image is a distinctly different interpretation.

Range: 0 to 100. Default: 0.

abstract landscape, mountains and fog --chaos 0

All four images will be variations on the same composition concept.

abstract landscape, mountains and fog --chaos 70

Each of the four images will take your prompt in a genuinely different direction — different composition, different mood, different interpretation.

When to Use --chaos

High chaos is useful when you're exploring a concept and don't know exactly what you want. It's like asking Midjourney to brainstorm four different ideas rather than four variations of one idea.

Low chaos (or leaving it at default) is better when you've figured out a composition you like and want four consistent variations to choose between.

I find --chaos above 40 usually produces at least one image in the grid that's unusable — something weirdly composed or off-topic. Below 30 gives you variation without losing coherence. That 20–30 range is genuinely useful for creative exploration.

--weird (or --w)

If --chaos creates variation in how your prompt is interpreted, --weird pushes the content itself into strange territory. It introduces unusual or unexpected elements — dreamlike logic, glitchy aesthetics, unexpected combinations.

Range: 0 to 3000. Default: 0.

portrait of a woman in a garden --weird 0

Standard, beautiful portrait.

portrait of a woman in a garden --weird 500

She might still be in a garden, but something will be off. The flowers might be made of something unexpected. The light might be wrong in an interesting way.

portrait of a woman in a garden --weird 1500

Genuinely strange. Might still have a face, might have a garden, but the relationship between subject and environment is now dreamlike.

My honest experience with --weird: It's one of those parameters that sounds exciting and mostly produces stuff that's hard to use. Below 300, the effect is subtle enough to be interesting without being confusing. Above 800, you're mostly in glitch-art territory that's hard to control. The sweet spot I've found is 200–400 for "interesting but not broken."

--quality (or --q)

Controls the amount of rendering detail. Higher quality = more processing time = more detail.

For final, high-resolution images you're going to use publicly, --q 2 is worth the extra generation time. For rapid iteration and testing prompts, --q 0.5 or --q 1 gets you results faster.

--no (Negative Prompting)

--no tells Midjourney what to exclude from the image. It's one of the most useful parameters and beginners often miss it entirely.

forest landscape, autumn colors --no people, cars, buildings, telephone poles

food photography, chocolate cake --no text, watermark, blurry background

portrait of a woman --no sunglasses, hat, jewelry

The --no parameter accepts comma-separated lists. It's not as granular as Stable Diffusion's negative prompt system, but it does work for excluding obvious, high-level elements. For a deeper exploration of what to exclude and why, the Midjourney Negative Prompts guide on this site covers the technique in detail.

--tile

Generates a seamless repeating tile pattern. Incredibly useful for textures, wallpapers, and design assets.

vintage botanical illustration pattern, leaves and flowers --tile --ar 1:1

geometric art deco pattern, gold and black --tile --ar 1:1

watercolor floral pattern, soft pastels --tile --ar 1:1

--tile works best at square aspect ratios and with patterns that have natural repetition in the subject matter. Abstract patterns and botanical illustrations work extremely well.

--seed

Every image Midjourney generates is based on a random seed number. If you use the same seed with the same prompt, you get the same image. This matters for consistency across a project.

To find a seed from a previous image: React to it with the 🌱 emoji in Discord, and the bot will send you the seed number.

portrait of a woman, red hair, studio lighting --seed 742891

Using seeds lets you iterate on a specific image direction — change one part of the prompt while keeping everything else the same.

Combining Parameters Effectively

The real skill is knowing how parameters work together. Here are practical combinations:

Photorealistic photography:

[subject], [camera reference], [lighting] --ar 3:2 --style raw --s 75

High-quality concept art:

[scene description], concept art, highly detailed --ar 16:9 --s 300 --q 2

Creative exploration:

[concept] --ar 1:1 --chaos 25 --s 200

Strange and experimental:

[subject] --weird 400 --chaos 15 --s 150

Seamless texture:

[pattern description] --tile --ar 1:1 --s 100

The Prompt Engineering course has a module specifically on parameter combinations and how they interact, which goes deeper than this guide can cover. The Prompt Cheatsheet has a parameters quick-reference section.

If you're using other AI image tools alongside Midjourney, the AI Image Generation section covers how parameters compare across different platforms, and the Free AI Tools section covers which free tools let you experiment with similar controls.

Quick Parameter Defaults Reference