Midjourney: What It Is, How Designers Use It, and What Rights You Get

What does it do and how is it used?

Midjourney takes a prompt and returns several variants of the image, from which you can request more variations or an upscaled version. Its default style is strong: it imposes an aesthetic even when you did not ask for one. That is why it helps to write the prompt with intent, specifying lighting, framing, palette, and references.

References are the most useful part for design. You can upload an image from your brand or from an inspiration board and ask the result to keep its palette or texture. These options are controlled with prompt parameters, and which ones are available depends on the model version: check the current documentation.

Why does it matter to a designer?

Because early exploration is cheap. Before you decide whether a travel app’s landing page should be bright or sober, you can generate thirty atmospheres in ten minutes and discard twenty. That does not replace judgment, but it speeds up the conversation with the team.

Midjourney is one piece of the workflow with tokens and references: the image helps fix intent, and the tokens translate it into values the code understands. If the image says “warm and calm” but the color token is a saturated orange, something was lost in translation.

What plans and rights does it include?

Midjourney is sold as a paid monthly subscription, with plans that differ in price, fast GPU hours and features such as Stealth mode. There are no figures here because they change: check the official pricing page and the Comparing Midjourney Plans article in its documentation before budgeting.

What affects a team most is not the price but what the terms imply. This is what you should review:

The third-party guide dates from February 2026 and the terms may have changed since. Read the official terms of service before delivering material to a client.

The most delicate point is visibility: if you work on a confidential project, do not use a plan without Stealth or a shared channel.

When should you NOT use it?

• When you need readable text inside the image, such as a logo or a mockup with final copy.
• When the result must be identical across sessions: consistency of characters or brand is hard to guarantee.
• When confidential material cannot be public by default.
• When a client requires traceable rights for each image: review their contract first.

Example

Illustrative example: a designer prepares a proposal for a meditation app. She uploads three brand photos as references and asks for four atmospheres with morning light and a palette of warm grays. She picks one, uses it as the background of the presentation, and, for the interface, defines color tokens from that palette in a file the team can read.

What she learned: the image aligned the client in one meeting, and the interface was built with real components, not with the image.

How do you write a prompt for a moodboard?

A useful prompt for exploration has five parts, in this order:

1. Subject: what appears, for example “a ceramic cup on a wooden table.”
2. Atmosphere: the emotion you want, such as “calm” or “morning energy.”
3. Light and framing: “soft side light,” “overhead shot,” “close-up.”
4. Palette: three or four colors or tones, not a list of ten.
5. Reference: an image from your brand or from inspiration, with its weight defined.

When the result drifts from what you wanted, change one part at a time. If you change five things, you will not know which one caused the shift. Save the prompt of every variant you like: without it, you cannot reproduce the direction or explain it to a client.

How do you move from image to tokens?

An image is not a design system, but it can feed one. A simple process:

1. Choose the image that best represents the direction.
2. Extract five dominant colors with a sampling tool and note their proportions.
3. Turn those colors into semantic tokens, such as color.surface.warm or color.text.primary, not color names.
4. Check text contrast against each background before you accept the palette.
5. Record which image gave rise to each token, so you can return to it.

That way the image fixes the intent and the code receives verifiable values. It prevents the visual direction from staying in a presentation and never reaching the interface.

What are the alternatives?

Midjourney is not the only image generator. To choose between several, compare four things: what privacy each plan offers, what license it grants over what you generate, how much control you get with reference images, and whether you can keep a coherent style across pieces. For the full picture read image generation for design, and for the specific use in visual direction, AI moodboards and consistent AI illustrations.

Common mistakes

• Using a generated image as part of the final interface, with text baked in.
• Uploading confidential material to a plan or channel where it is public by default.
• Assuming a generated image is free to use without reading the current terms.
• Taking the first variant without exploring, or exploring without a clear brief.
• Moving from image to code by eye, without turning the palette into tokens.

Sources

• Midjourney License Explained: Commercial Use, Ownership, and the Fine Print (2026), third-party guide on license, visibility and commercial use
• For current prices, plans and terms (the official pages block automated reading, which is why no figures are cited): pricing, plan comparison and terms of service