← All guides
Guides

Midjourney prompt library for designers

Image prompts are not text prompts with pictures. Your Midjourney library needs its own structure — weights, ratios, styles, version flags.

Published April 28, 20267 min read

A Midjourney prompt library looks almost nothing like a ChatGPT prompt library. The syntax is different, the iteration loop is different, the variables are different. If you are trying to reuse a generic text-prompt library for Midjourney, you will feel the mismatch immediately. Here is how to build one that actually fits the tool.

Why Midjourney prompts need their own library

Text prompts and Midjourney prompts have different shapes. ChatGPT prompts are sentences with intent. Midjourney prompts are comma-separated token lists with suffix parameters: --ar 16:9, --stylize 250, --v 6.1. They reward precision on style tokens and punish verbose explanation.

The four parts of every Midjourney prompt

Subject, style, composition, parameters. Every serviceable Midjourney template has slots for all four. Subject is what is in the frame. Style is the visual reference — editorial photography, Studio Ghibli, architectural render. Composition is framing — wide shot, overhead, macro. Parameters are the suffix flags: aspect ratio, stylize weight, version.

Template structure that works

Template: "{{subject}}, {{style_reference}}, {{composition}}, {{lighting}}, {{mood}} --ar {{aspect_ratio}} --stylize {{stylize}} --v {{version}}"

Example fill: "A young woman reading in a cafe, editorial photography, medium shot, soft window light, contemplative --ar 3:2 --stylize 250 --v 6.1"

Seven variables sounds like a lot. In practice, half of them are pre-filled from your last use — which is the whole point of a library with variable history.

Style tokens: the reusable gold

Most of the value in a Midjourney library is in style tokens. "Editorial photography, muted palette, Leica 35mm" is a style spine you will reuse across dozens of subjects. Pull these out as their own category. Reuse them as building blocks instead of rewriting them for every prompt.

Version flags drift — make them a variable

Midjourney ships new versions often. Hardcoding --v 6 in 40 prompts means you rewrite 40 prompts the day v7 lands. Mark it as {{version}} and update once when you switch.

Aspect ratio deserves its own variable

Almost every prompt is run at two or three aspect ratios (16:9 for web, 3:2 for editorial, 1:1 for social). Having {{aspect_ratio}} as a variable makes generating the three variations a ten-second task instead of a cut-and-paste chore.

Negative prompts and seed management

If you are iterating on a scene, you want --no variables for common rejects (text, watermark, low contrast) and a {{seed}} variable for reproducibility. A template that bakes these in saves the "why does the hand look like that" spiral.

Where this lives

Most generic prompt libraries do not handle Midjourney well because the {{variables}} approach was built for text prompts. TextDeck handles both — the variable system is neutral, so a template with "subject, style, composition, --ar {{aspect_ratio}}" works exactly the same as a text prompt. Same hotkey (⇧⌘P), same fill-in flow, same fast paste into the Midjourney Discord or web app.

Try the workflow yourself