What does --chaos do in Midjourney?

The --chaos parameter (range 0 to 100, default 0) controls how different the four images in a single generation are from each other. Low chaos produces four similar, reliable variations on one idea; high chaos produces four wildly different interpretations of the same prompt.

What is the default chaos value?

The default is 0, meaning the four results are coherent variations on a single interpretation. Many users never change it. Raise --chaos when you want the model to surprise you with divergent directions rather than refining a single concept.

When should I use high chaos?

Use high chaos (50 to 100) at the start of a project when you are exploring directions and want maximum variety from one prompt — moodboarding, ideation, or breaking out of a creative rut. It trades consistency for surprise.

When should I use low chaos?

Use low chaos (0 to 25) when you have a clear vision and want all four images to be close, predictable takes you can pick from and upscale. It is the right choice for production work, brand assets, and refining a known concept.

How is chaos different from stylize?

They control different things. --chaos controls how varied the four results are from each other, while --stylize controls how strongly Midjourney applies its own aesthetic to each one. You can combine them — for example high chaos with low stylize gives varied but literal interpretations.

What is the Midjourney Chaos Parameter Guide?

Interactive guide to Midjourney's --chaos parameter. Move the slider from 0 to 100 to see how much variation and unpredictability each level adds across the four-image grid, with use-case recommendations and copy-ready --chaos snippets. It runs free in your browser on Gera Tools, with nothing uploaded.

Midjourney Chaos Parameter Guide

Name: Midjourney Chaos Parameter Guide
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

Get one useful tool a week

Like this tool? Enter your email and we'll send you one genuinely useful Gera tool a week — plus a link to come back to this one. No spam, one-click unsubscribe any time.

Midjourney’s --chaos parameter decides how adventurous a single generation is. At 0, the four images in the grid are coherent variations on one idea. At 100, they can be four completely different interpretations of the same words. This guide lets you scrub the value and see exactly what to expect — plus when each level is the right tool.

How it works

--chaos (often abbreviated --c) takes an integer from 0 to 100, defaulting to 0. The slider here maps the range into practical bands:

0–10 — Coherent. Four close variations on a single interpretation. Predictable and reliable.
11–25 — Gentle variety. Small divergences in pose, framing, or palette. Still on-theme.
26–50 — Exploratory. Meaningfully different takes that share the core idea. A good default for brainstorming with some safety.
51–75 — Divergent. The four images branch into distinct concepts. High variety, lower consistency.
76–100 — Wild. Maximum unpredictability. The model treats the prompt as a loose springboard.

Higher chaos is a discovery tool: you give up the ability to predict the output in exchange for a wider spread of ideas from a single prompt and a single set of GPU minutes.

Tips and notes

Start high, then commit. Use --chaos 50–100 early to find a direction, note the seed of the image you like, then drop chaos back to 0 to refine it.
Chaos and stylize are independent. Combine them deliberately — high chaos + low stylize gives varied but literal results; low chaos + high stylize gives consistent but heavily stylised ones.
It does not cost extra. Like stylize, chaos changes the look, not the render time or GPU usage. Experiment freely.
Pair with --no for control. If high chaos keeps surfacing unwanted elements, add a --no exclusion list to keep the variety on-brief.

Chaos in practice: a creative workflow

The most effective use of --chaos is not to pick a single value and stick with it, but to use different values at different stages of a project:

Stage 1 — Discovery (--chaos 75–100). Start with a short prompt and maximum chaos to see what directions are even possible. You might be surprised by how the model interprets an abstract concept at high chaos; unexpected directions can be more interesting than what you originally imagined.

Stage 2 — Narrowing (--chaos 25–50). Once you have identified an interesting direction from the chaos run, refine the prompt to reflect that direction and reduce chaos. Now you are exploring variations within the lane you like, rather than across all possible lanes.

Stage 3 — Execution (--chaos 0–10). Lock in a specific concept and drop chaos to near zero. You want four close interpretations you can pick from to upscale. Note the seed of the one you like so you can reproduce or vary it precisely.

How chaos interacts with seeds

At --chaos 0, passing the same seed (--seed N) to the same prompt reliably reproduces the same image. As chaos rises, the seed still influences the generation, but the high variance means small changes produce more divergent results. This means seeds are most useful as a reproducibility tool at low chaos, and less useful as a fine-tuning tool at high chaos where the output space is intentionally wide.