Midjourney Tutorial: From Prompt to First Image

People searching for a Midjourney tutorial rarely want another “what is the model” explainer. They want a prompt to become a usable first image. The fastest path is not stacking adjectives from a blank box. It is: copy a complete template → set --ar and speed on the generate page → run once → change only one variable and run again. Walk that sequence below.

Copy from the prompt library instead of starting blank

A blank field invites “a beautiful woman” or “cyber city,” which leaves composition to chance. Library entries already order subject, setting, light, and lens, and they include parameters such as --ar. For your first image, open the Prompt Library, pick the closest genre (portrait, product, landscape, anime), paste the whole block, and replace only one noun.

Starter skeleton you can paste:

cyber courier in a rainy neon alley, teal and magenta rim light, 85mm cinematic portrait --ar 3:4 --stylize 200

Change only “cyber courier” to your character. Leave light and lens alone so you can see which clause Midjourney actually followed. Parameter spelling follows the official docs; this page maps those flags onto the generate-page controls.

Set —ar and speed on the generate page, then run

Open the Generate page, paste the prompt, and check two things before you start:

1. Aspect --ar: portraits 3:4 or 9:16, wide scenes 16:9, catalog squares 1:1. A wrong ratio crops faces and product edges later.
2. Speed: use Fast the first time. Relax is for queued batches; Turbo is faster and spendier. Confirm the picture direction before you chase throughput.

Wait for the grid. The first image is not a final deliverable. Confirm three things: the subject is present, the light is roughly right, and the crop did not clip a person. Only then iterate.

Fast, Relax, and Niji: pick a line for image one

The same prompt on the wrong model line will look like a different brief:

• Standard model + Fast: default for portraits, products, interiors, and landscapes.
• Niji: character sheets, manga panels, illustrated color planes. See the Niji 7 tutorial.
• Relax: use after the skeleton is stable and you need a dozen variants.
• Turbo: rush concept frames. It is not a quality switch.

Stacking cinematic 85mm, photoreal skin on Niji fights the line work. Opening Niji on a product shot cartoonizes edges. Lock one line for the first image and keep the series on that line.

When the result is close, change only one thing

If the grid is already near the brief, do not rewrite the whole prompt. Change in this order:

1. Subject right, light wrong → edit only the lighting clause (backlight / softbox / overcast).
2. Crop right, mood wrong → nudge --stylize (commercial work can start at 150–250).
3. Local errors (hands, type, clutter) → Vary Region instead of a new run.
4. Completely wrong direction → swap to another library template, do not pile words onto a bad skeleton.

The writing breakdown lives in How to Write Better Midjourney Prompts. This tutorial only gets you through image one. Parameter lessons continue in the beginner intro and Vary Region.

Turn image one into a reusable workflow

After one clean run, save the prompt as a template: skeleton + default --ar + default speed + model line (standard or Niji). Next time, replace one variable in the subject or setting. Series stay steadier, and you will not need to re-search a Midjourney tutorial every session.

Go back to the Generate page with the same template and run two variants: light-only vs --ar-only. If you can say what you changed and where the picture moved, you have actually finished this tutorial.