Midjourney Prompt Engineering Guide

A Prompt in basics – is a short text phrase that the Midjourney interprets to produce defined in prompt .
The Midjourney Bot breaks down the words and phrases in a prompt into smaller pieces, called tokens, that are compared to its training data and then used to generate an image. A well-crafted prompt can help make unique and exciting images.

Basic Prompt

A basic prompt can be as simple as a single word, phrase or even emoji.

What’s very important here even for basic prompts:
The Midjourney Bot works best with simple, short sentences that describe what you want to see. Avoid long lists of requests and instructions.
Instead of: Show me a picture of lots of blooming California poppies, make them bright, vibrant orange, and draw them in an illustrated style with colored pencils Try: Bright orange California poppies drawn with colored pencils

More Advanced Prompt

More advanced prompts can include one or more image URLs, multiple text phrases, and one or more parameters

Where we can define:

Text Prompt

The text description of what image you want to generate. See below for prompting information and tips. Well-written prompts help generate amazing images.

Image Prompts

Image URLs can be added to a prompt to influence the style and content of the finished result. Image URLs always go at the front of a prompt.

Parameters

Parameters change how an image generates. Parameters can change aspect ratios, models, upscalers, and lots more. Parameters go at the end of the prompt.

PROMPTS PARAMETERS

Aspect Ratio

The aspect ratio parameter (–-aspect or –-ar) is used to alter the aspect ratio of a generated image. The aspect ratio is the ratio of the width to the height of an image. For example, an aspect ratio of 1:1 means that the image is a square, while an aspect ratio of 16:9 means that the image is wider than it is tall. The default aspect ratio for Midjourney AI is 1:1.

To use the aspect ratio parameter, you need to specify the width and height values after the parameter name, separated by a colon. For example, –-aspect 2:1 means that the image will have twice as much width as height. You can use any values for the aspect ratio, but keep in mind that values greater than 2:1 are experimental and may produce unpredictable results.

There’s also update with landsace, portrait or movie ratio, following: —l for —ar 3:2 (landscape) —p for —ar 2:3 (portrait) —m for —ar 16:9 (movie, or 21:9)

Here are some examples of using the aspect ratio parameter:

a snowy mountain –-aspect 2:1
a cute cat –-aspect 1:2
a colorful abstract painting –-aspect 3:4

Chaos

The chaos parameter (–-chaos) dictates the diversity of the results. It accepts values from 0 to 100, where higher values lead to more unexpected and unusual outputs, while lower values lead to more coherent and realistic outputs. The default value for chaos is 0.

To use the chaos parameter, you need to specify a value after the parameter name, separated by a space or a dash. For example, –-chaos 50 means that the image will have a moderate level of variation and randomness.

Here are some examples of using the chaos parameter:

a dragon –-chaos 0
a dragon –-chaos 50
a dragon –-chaos 100

Image Weight

The image weight parameter (–-iw) sets the image prompt weight relative to the text weight. It accepts values from 0 to 2, where higher values mean that the image prompt is more important than the text prompt, while lower values mean that the text prompt is more important than the image prompt. The default value for image weight is 1.

To use the image weight parameter, you need to specify a value after the parameter name, separated by a space or a dash. For example, –-iw 2 means that the image prompt will have twice as much influence as the text prompt.

Here are some examples of using the image weight parameter:

a forest with a waterfall [image of a forest] –-iw 0
a forest with a waterfall [image of a forest] –-iw 1
a forest with a waterfall [image of a forest] –-iw 2

No

The no parameter (–-no) enables negative prompting. It allows you to remove an element from the image by specifying what you don’t want to see in the output. For example, –-no plants would aim to eliminate plants from the image.

To use the no parameter, you need to specify what you want to remove after the parameter name, separated by a space or a dash. For example, –-no sky means that the image will not have any sky in it.

Here are some examples of using the no parameter:

a beach –-no water
a cityscape –-no buildings
a portrait –-no eyes

Quality

The quality parameter (–-quality or –-q) determines the amount of rendering quality time you wish to use. It accepts values of 0.25, 0.5, or 1, where higher values consume more GPU minutes (the unit of measurement for Midjourney AI usage), while lower values consume less GPU minutes. The default value for quality is 1.

To use the quality parameter, you need to specify a value after the parameter name, separated by a space or a dash. For example, –-quality 0.5 means that the image will use half as much GPU minutes as normal.

Here are some examples of using the quality parameter:

a sunset over the ocean –-quality 0.25
a sunset over the ocean –-quality 0.5
a sunset over the ocean –-quality 1

Seed

The seed parameter (–-seed) is used by the Midjourney AI bot to create a field of visual noise, akin to TV static, as a starting point for generating initial image grids. While seed numbers are randomly generated for each image, they can be specified using the seed parameter. It accepts integers between 0 and 4294967295. Utilizing the same seed number and prompt results in similar ending images.

To use the seed parameter, you need to specify a value after the parameter name, separated by a space or a dash. For example, –-seed 123456 means that the image will use the same seed number as another image with the same prompt and seed.

Here are some examples of using the seed parameter:

a haunted house –-seed 666
a haunted house –-seed 777
a haunted house –-seed 888

Stop

The stop parameter (–-stop) completes a job partway through the process. It accepts values between 10 and 100, where lower values mean that the image will be less detailed and more blurry, while higher values mean that the image will be more detailed and less blurry. The default value for stop is 100.

To use the stop parameter, you need to specify a value after the parameter name, separated by a space or a dash. For example, –-stop 50 means that the image will be half as detailed as normal.

Here are some examples of using the stop parameter:

a flower –-stop 10
a flower –-stop 50
a flower –-stop 100

Style

The style parameter (–-style) allows for switching between different versions of the Midjourney and Niji models. Midjourney Team regularly updates its models to enhance their performance, coherence, and quality. The default version for Midjourney AI is version 5.1.

To use the style parameter, you need to specify a value after the parameter name, separated by a space or a dash. For example, –-style raw means that the image will use Midjourney model version 5.1.

Here are some examples of using the style parameter:

a unicorn –-style raw
a unicorn –-style cute
a unicorn –-style expressive

Stylize

The stylize parameter (–-stylize or –-s) influences the extent to which Midjourney’s default aesthetic style is applied to jobs. It accepts values from 0 to 1000, where higher values mean that the image will have more stylization and abstraction, while lower values mean that the image will have less stylization and abstraction. The default value for stylize is 100.

To use the stylize parameter, you need to specify a value after the parameter name, separated by a space or a dash. For example, –-stylize 500 means that the image will have a moderate level of stylization.

Here are some examples of using the stylize parameter:

a landscape –-stylize 0
a landscape –-stylize 500
a landscape –-stylize 1000

Tile

The tile parameter (–-tile) produces images that can serve as repeating tiles, thereby creating seamless patterns.

To use the tile parameter, you just need to add it to the end of your prompt without any value. For example, –-tile means that the image will be suitable for tiling.

Here are some examples of using the tile parameter:

a geometric pattern –-tile
a floral design –-tile
a starry night –-tile

PROMPTING TIPS & TRICKS

Word Choice

Word choice matters. More specific synonyms work better in many circumstances. Instead of big, try tiny, huge, gigantic, enormous, or immense.

Plural words and Collective Nouns

Plural words leave a lot to chance. Try specific numbers. “Three cats” is more specific than “cats.” Collective nouns also work, “flock of birds” instead of “birds.”

Focus on What you Want

It is better to describe what you want instead of what you don’t want. If you ask for a party with “no cake,” your image will probably include a cake. To ensure an object is not in the final image, try advanced prompting using thee --no parameter.

Prompt Length and Details

Prompts can be simple. A single word or emoji will work. However, short prompts rely on Midjourney’s default style, allowing it to fill in any unspecified details creatively. Include any element that is important to you in your prompt. Fewer details means more variety but less control.

Try to be clear about any context or details that are important to you. Think about:

Subject: person, animal, character, location, object
Medium: photo, painting, illustration, sculpture, doodle, tapestry
Environment: indoors, outdoors, on the moon, underwater, in the city
Lighting: soft, ambient, overcast, neon, studio lights
Color: vibrant, muted, bright, monochromatic, colorful, black and white, pastel
Mood: sedate, calm, raucous, energetic
Composition: portrait, headshot, closeup, birds-eye view