GALAXY OF TOOLS

AI SECTOR

Midjourney Prompt Generator

Build Midjourney prompts with checked parameters, aspect ratios, permutations and multi-prompt weights.

What are you making?
The image
Short, concrete phrases work best. {red, green} makes one prompt per option; text in "double quotes" is drawn in the image.
Describe a look instead of naming a living artist.
Short words in Latin letters render best (V6 and later).
Multi-prompt with weights (V6, V6.1, Niji 6)
No weight means 1. A negative weight removes something; the weights must add up to more than zero.
The video
On midjourney.com, drop the image into Starting Frame instead and leave this empty.
Describe what moves; the image already shows what is there.
Motion
Parameters
Each version accepts different parameters — the checklist knows which.
Aspect ratio (--ar)
Whole numbers only — 139:100, not 1.39:1. See the aspect ratio calculator.
0–1000
0–100
0–3000
Comma-separated. Each word is read on its own.
References and personalization
0–1000
V7: 1–6 (default 6). V6: 1–4 (default 4).
Placed at the start of the prompt.
0–3 (Niji 7: 0–2)
1–1000
Replaces Omni Reference and Character Reference on V8.
Advanced
0–4294967295
V7: 1, 2, 4 · V6: 0.5, 1, 2
2–40 by plan
Permutations and repeats: 4 / 10 / 40.
Typed as-is at the end. The checklist reads them too.

The prompt is put together in your browser — nothing is sent to Midjourney or anywhere else. Share links keep it after the #, which is never sent to a server; favourites stay in this browser.

A Midjourney prompt generator that checks every parameter

This Midjourney prompt generator turns a subject, a medium and a few descriptive choices — lighting, colour, mood, composition, lens — into one clean prompt line, then adds parameters such as --ar, --s, --c, --w and --no in the right syntax. As you type, a checklist tests the values against the ranges in Midjourney’s own documentation and against the model version you picked.

That matters because versions differ: V8.1 and V8.2 have no --q, Omni Reference works only on V7, multi-prompts with :: work only on V6-era models, and HD images stop at 4:1. The page also previews permutations — {red, green} groups — so you see every prompt, and the job count, before Midjourney asks you to confirm.

For other models, the AI image prompt generator writes the same idea for GPT Image, Stable Diffusion, Flux and Ideogram, and the AI video prompt generator plans video shots.

How to write a Midjourney prompt here

  1. Pick an example under Start from an example, or choose Image or Video from an image.
  2. Describe the Subject and scene, then add a Medium and a Movement, era or technique rather than an artist’s name.
  3. Choose a Model version and an Aspect ratio; the box below shows the shape and an estimated size in pixels.
  4. Set Stylize, Chaos and Weird, list anything to Leave out, and open References or Advanced for style codes, seeds, repeats and your plan.
  5. Fix anything the checklist marks, then press Copy — or copy every expanded prompt from the Expanded tab.

Midjourney Prompt Generator features

Version-aware checks

Ranges and compatibility for V8.2 (the default since 24 July 2026), V8.1, V7, V6.1, V6, Niji 7 and Niji 6.

Permutation preview

Expands {a, b} groups — nested ones and \, escaped commas too — and counts prompts and jobs against your plan’s 4, 10 or 40 limit.

Aspect ratio shapes

Seven presets plus custom ratios, with a shape preview, a 14:1 limit check and the HD 4:1 rule.

Multi-prompt weights

Builds space::2 ship-style prompts with the documented spacing and checks that the weights add up to more than zero.

Wording checks

Flags repeated words, “no …” phrases that belong in --no, single-quoted text, artist names and unknown or legacy parameters.

Video prompts

Image-to-video lines with --motion, --raw, --loop, --end and --bs.

Current Midjourney parameters and ranges

Checked against docs.midjourney.com on 14 October 2026. Defaults are what you get when the parameter is left out.

ParameterRangeDefaultNotes
--arwhole numbers, up to 14:11:1HD images up to 4:1
--s (stylize)0–1000100Also scales Personalization
--c (chaos)0–1000Variety between images
--w (weird)0–30000Not fully compatible with seeds
--seed0–4294967295random“99% identical” on V8
--qV7: 1, 2, 4 · V6: 0.5, 1, 21Not on V8.1/V8.2
--sw / --sv0–1000 / V7 1–6, V6 1–4100 / 6 (V7)Need --sref
--iw0–3 (Niji 7: 0–2)1Needs an image prompt
--oref / --owone image / 1–1000— / 100V7 only
--r2–40 (Basic 4, Standard 10)—Fast or Turbo only

Also supported: --raw, --tile, --no, --sref, --p, --draft (V7 and V8; V8 drafts on the website only), --hd/--sd (V8), --edit (V8.1 and V8.2, up to four images), --fast/--relax/--turbo (no Turbo on V8) and --stealth/--public. Typed legacy parameters such as --cref, --stop or --style are flagged; --exp appears in Midjourney’s version chart without a documented range, so it is passed through unchecked.

Worked example

The first example combines a medium, a movement instead of an artist, lighting, colour, mood and framing with a 3:2 ratio, a stylize value and two exclusions:

watercolour and ink illustration of a lighthouse keeper’s cottage on a windswept sea cliff, Art Nouveau style, golden hour light, muted teal and amber palette, calm, nostalgic mood, wide shot, low horizon --ar 3:2 --s 250 --no people, text

All five key checks pass. Load “Permutations: colours × places” and the line a {red, green, yellow} bird in the {jungle, desert} expands to six prompts — six jobs — which fits the Standard plan’s limit of ten:

colored pencil illustration of a red bird in the jungle --ar 1:1
colored pencil illustration of a red bird in the desert --ar 1:1
colored pencil illustration of a green bird in the jungle --ar 1:1
colored pencil illustration of a green bird in the desert --ar 1:1
colored pencil illustration of a yellow bird in the jungle --ar 1:1
colored pencil illustration of a yellow bird in the desert --ar 1:1

Prompting tips from Midjourney’s guides

  • Short and concrete — “colored pencil illustration of bright orange California poppies” beats a long list of instructions.
  • Numbers and collective nouns — “three cats” or “a flock of birds” is clearer than a plural.
  • Say what you want — a party with “no cake” may still get a cake; put exclusions in --no, which reads every word on its own.
  • Text in “double quotes” — short Latin-alphabet words render best; lower Stylize or use Raw if the letters come out wrong.
  • Seeds are for testing — they fix the starting noise, not a style.

Midjourney Prompt Generator FAQ

What is the default Midjourney version in October 2026?

V8.2, the default since 24 July 2026. V8.1, V7, V6.1, V6, Niji 7 and Niji 6 can still be chosen with --v or --niji. This Midjourney prompt generator leaves --v out unless you pick a version, so your account default applies.

What is the maximum stylize value?

1000. Stylize runs from 0 to 1000 with a default of 100; the checklist accepts 1000 and flags 1001, negative numbers and decimals.

Why does my prompt say --q is not allowed?

Midjourney’s version chart lists no quality parameter for V8.1 and V8.2. Use --q only with V7 (1, 2 or 4) or V6 and V6.1 (0.5, 1 or 2).

How many permutations can I run at once?

Basic allows 4 prompts from one permutation prompt, Standard 10, Pro and Mega 40, and only in Fast or Turbo Mode. Each prompt uses Fast GPU time, and --r multiplies the jobs.

Can I copy an artist’s style?

The page encourages describing movements, eras, media and techniques instead — such as Art Nouveau, risograph or 1930s travel poster — and warns when a prompt names an artist.

Are the pixel sizes exact?

No. Midjourney documents 1024 × 1024 for SD and 2048 × 2048 for HD on V8.2; other ratios are estimated at the same area and marked as estimates.

Does this page connect to Midjourney?

No. It builds the text in your browser; you paste it into midjourney.com or Discord.

You might also need