Generate Midjourney prompts from a mood board

My prompts were hit or miss. Now ChatGPT translates the brief into Midjourney language.

The workflow

1. Describe the mood and references
2. Ask ChatGPT for 5 Midjourney prompts
3. Include style, lighting and aspect ratio
4. Generate and pick the best

Result: Prompts that match the brief on the first try more often.

The prompt

I want images that feel [MOOD]. Write 5 Midjourney prompts with subject, style, lighting, and --ar. No text in the image.

A note on what follows. The workflow, the prompt and the result above come from the member. The notes below (why the prompt works, what to watch for, how to adapt it) were written with AI help.

Why this prompt works

Listing the four things I want, subject, style, lighting, aspect ratio, is what stopped me getting five variations of the same photo. Before that I asked for prompts in a mood and got five near-identical lines with the adjectives shuffled. Naming lighting separately was the biggest single change, because lighting is most of what makes an image feel like something, and left to itself it writes "soft lighting" every time. "No text in the image" is there because Midjourney kept adding garbled signage. Five is deliberate too. Three does not give me enough to choose from, ten and they start repeating. I still throw away three of the five, which I think is normal.

What to watch for

• Invented parameters. It will write flags that do not exist in the version you are on, and Midjourney either ignores them or errors.
• Aspect ratios that do not match your actual placement. It defaults to 16:9 for things I need square.
• Style words that are really just other brands. "Editorial" and "cinematic" are fine, a living photographer's name is a different conversation.
• The five prompts converge on the same time of day. Ask for a mood and you get golden hour five times.

How to adapt it

• Force variety with Each prompt must use a different time of day and a different distance from the subject.
• For a brand palette, add Use only muted sage, bone and warm grey. Vague colour instructions get ignored.
• If you are iterating on one image, add Keep the subject and composition identical, vary only the lighting.

What good output looks like

I read the five as a set. If I could describe them all in one sentence, the set has failed and I ask again with the variety constraint. Each prompt should name a subject I can picture before I read the style words. Long strings of adjectives with no subject produce mush. And the aspect ratio should match where the image is going, which sounds obvious and is the thing I forget most.