Artificial Lexicon
Tool notes

Using the walls in Midjourney

Midjourney rewards short prompts. The walls give you the words. This page shows how to put them together.

Prompt anatomy

A Midjourney prompt that behaves has this shape: subject, style, mood, details, parameters. Front-load what matters. Words earlier in the prompt carry more weight.

shape/imagine [subject doing something], [one style word], [one mood word], [2–3 details] --ar 3:2 --stylize 200

Build the middle three slots from the walls: style from Styles & Movements, mood from Lighting & Atmosphere, details from Materials or Composition. The tray on each wall assembles the comma list for you.

The parameters that matter

  • --ar: aspect ratio. Decide it first. Composition words behave differently in portrait and wide frames.
  • --stylize (0–1000): how much Midjourney imposes its own taste. Keep it low when your style words are doing the work, or they'll fight.
  • --chaos (0–100): variety across the four candidates. Raise it when exploring a wall, drop it when refining.
  • --no: the negative prompt. It works better than writing "without" in prose.
💡 Wall words + high --stylize is the most common way prompts go mushy. If the output ignores your style chip, halve the stylize value before changing words.

A wall-first workflow

  1. Write the subject plainly, no adjectives.
  2. Open one wall, pick one chip, generate. See what that word alone does.
  3. Add a second wall's chip. Style plus lighting gives the most leverage.
  4. Only then touch parameters.

One variable at a time. Otherwise you never learn which word earned its place. That's the whole trick, and it's why the walls are sorted by job instead of alphabet.

Using Nano Banana or GPT-Image instead? The words transfer; the chat image models guide covers the dialect. First time prompting? Start Here walks the first prompt end to end. Want the long version? 15 Tips to Not Suck at MidJourney covers iteration habits, seeds, and remixing.