Guide to Midjourney

Welcome to our guide about Midjourney, specially developed for Immersive Lab partners and our TETRA projects. In this manual, we share our expertise on how to effectively use Midjourney in your pre-production process, based on our practical experience and proven workflows.

While we share our own experiences and best practices in this guide, we definitely encourage you to consult other sources! There are countless passionate Midjourney experts on YouTube sharing fantastic tutorials and tips. From beginner tips to advanced techniques – you’ll find a wealth of information. Dive into that YouTube rabbit hole and let yourself be inspired by the many creative souls sharing their knowledge! 😊

Midjourney also has a very extensive FAQ that you can always consult. We will regularly refer to it here as well.

By combining different perspectives and methods, you’ll develop your own style and workflow. Consider this guide as a springboard for further exploration of what Midjourney has to offer.

As a creative team at Immersive Lab, we’re fully immersed in the world of AI image generation! We haven’t just extensively tested Midjourney, but we’ve played with it, experimented with it, and ultimately embraced it in our daily work.

But we go even further – we mix and match different AI tools to achieve truly unique results. Think about the power of ChatGPT for perfecting prompts, Stable Diffusion for specific styles, and even our own trained ComfyUI models for extra special customization. And much more, which we’re working on here in the LAB.

Whether you’ve been working with AI tools for years or are just getting started – don’t worry! This guide is written with everyone in mind.

In the following sections, we’ll take you step by step through the process: from generating initial concepts to refining your final result. We share practical tips, concrete examples, and methods that are directly applicable to your own projects. Our focus is on maximizing the use of both ChatGPT and Midjourney for creating high-quality visual content.

Have fun prompting!

ChatGPT-4

ChatGPT, developed by OpenAI, is an advanced language model capable of conducting natural conversations and assisting you in generating ideas, writing content, and solving complex problems. Other language models like Microsoft’s Copilot, Jasper.AI, Monica, Claude, etc., apply similar techniques.
We will mainly focus on ChatGPT, as it is currently the most comprehensive model.

This guide dives deep into the possibilities of these AI tools:

Step 1A - Training chatgpt

The Brand
When training, it’s essential to clearly define the core of your brand. Think about the fundamental values, unique characteristics, and distinctive personality of your brand. What makes your brand special? How does it stand out in the market?

The tone of voice, visual identity, and communication style are key elements ChatGPT needs to understand to generate authentic content. Without this knowledge, the model may sound too “abstract and generic.” By carefully specifying these elements, you ensure that the AI model accurately represents your brand.

To achieve optimal results, it’s important to “feed” the model with concrete examples of your brand identity. Start by collecting strong, distinctive terms and expressions your target audience instantly recognizes. These keywords will form the foundation for consistent and recognizable brand communication that resonates with your audience.

Voice and Style
It’s crucial to define the tone and style that match your communication. Consider the following questions:

What’s the pace of your communication? Do you prefer short, punchy sentences or more elaborate text?
Do you use formal or informal language?
How technical can the language be?
What emotional tone do you want to convey (inspiring, professional, friendly, etc.)?

ChatGPT can be trained to apply these stylistic elements consistently across all generated content. It’s important to provide concrete examples of the desired language so the model can learn and replicate your unique voice.

As a marketing professional, you of course retain the freedom to adjust and refine ChatGPT’s output according to your expertise and creative vision. The model serves as a starting point that you can shape to meet the specific needs of each project.

Audience
Understanding your target audience is essential for effective AI-driven communication. When training ChatGPT, we need a detailed picture of, for example:

- Demographics: age, generation, location, education level, profession, etc.
- Psychographics: values, interests, lifestyle, etc.
- Communication preferences: which channels do they use, what type of content appeals to them?

By carefully documenting and inputting this information into ChatGPT, we can train the model to generate content that:

- Directly resonates with your audience’s world
- Uses the right tone and vocabulary that matches their level and expectations

The more detailed your audience definition, the better ChatGPT will be able to produce relevant and effective content that truly speaks to and activates your audience.

Step 1B - Training chatgpt

The Watt Method: The Key to Effective AI Communication

This method is designed to get the most out of your AI conversations. It’s a structured approach that helps you write clear, purposeful prompts that deliver the exact results you’re looking for. We use a practical version here, refined through many team experiments.

The WATT method consists of four essential components, which we will explore step-by-step:

W – Who: Defines the AI’s role and expertise
A – Background: Provides the context and relevant information
T – Task: Describes the specific objective
T – Tone: Specifies the desired style and mood

Below you’ll find a more detailed breakdown of this method:

WHO – Role & Task Setting

First, let’s clarify what you need help with and what role ChatGPT will take on. This helps us set clear expectations for the AI.

A few examples:

I’m a documentary filmmaker working on a film about X. Can you be my creative advisor and story editor?
I’m a director planning a new brand film project. Could you be my creative strategist and provide some concept ideas?
I’m designing a brand and need help structuring the brand story. Would you step into the role of brand strategist and give me feedback?

BACKGROUND – Details & Context

Provide relevant background information and context to help ChatGPT generate a more informed and nuanced response. This adds depth to the AI’s understanding of your task.

Examples:

The film is about a musician who overcomes adversity. Here are key details about their life and emotional arc.
The brand wants an ad about sustainability. Here’s background on their values, mission, and past campaigns.
The client is a non-profit promoting youth arts education. Here’s their website, past videos, and core messages we want to convey.

Describe the specific goal or desired outcome of the task. This tells ChatGPT what end result it should aim for.

Examples:

Please give 3 ways I could improve this story to make it more effective for the brand that hired me.
Review this transcript and identify key moments based on the brand’s story—6 essential plot points.
Suggest 3 new interesting angles the story could explore and highlight potential narrative hooks.

Provide specific guidelines or instructions to shape the tone, voice, style, and brand elements in ChatGPT’s output. This allows you to control the style and mood of the response.

Examples:

Make the tone feel intimate, authentic, and emotionally moving.
Incorporate the eco-friendly and community-focused voice of the brand.
Ensure the visual ideas have a raw, unscripted style suitable for a documentary with real people.

This method didn’t just appear “out of the blue”—it’s based on proven techniques. A similar English variant is the FRAME method.

Step 1C - Other methods for training Chatgpt

Besides the WATT and FRAME methods, several other popular frameworks exist to train ChatGPT effectively:

It’s important to remember that each method has its own strengths, but they are generally quite similar. You can combine elements from different methods to create an approach that best fits your specific needs. 💡 Pro tip: Experiment with different methods and keep track of which approach delivers the best results for your use cases.

CRISPE Method
A structured approach that stands for:

Capacity (role ChatGPT should take on)
Result (desired outcome)
Information (relevant context)
Style (tone and style)
Parameters (specific requirements)
Example (clarifying examples)

You can find an example below.

Learn more about CRISPE here:
CRISPE Method Explained

CARE Framework
A human-centered approach focusing on:

Context (situation and background)
Action (desired action)
Response (expected output)
Example (sample scenarios)

Example of a prompt using the CARE framework:

Context:
“I’m a marketing manager at a sustainable fashion brand. Our brand targets eco-conscious consumers aged 25–40 who value quality and sustainability.”
Action:
“Can you help me write a social media post for the launch of our new recycled denim collection?”
Response:
“I’d like a short, punchy post of about 3–4 sentences with a call to action.”
Example:
“Similar to our previous post about the organic cotton line, but now focusing on circularity and innovation.”

Start to DISCORD

To get started with Midjourney, you first need to create an account. The process begins at midjourney.com, where you click the “Sign up” button. This button takes you to a choice between Discord and Google. Choose Discord, where the real adventure begins.

Start at midjourney.com. Here you can choose between Discord or Google. For this guide, we’ll use Discord, as it is more stable and works better than the web version (which we’ll discuss later), which still has some minor bugs.

Discord

First, create a Discord account if you don’t have one yet.
Set up your own Discord server for a quieter workspace.

Follow these steps to set up your own server:

- Click the ‘+’ icon in Discord

Choose ‘Create Server’ and click on ‘Create My Own’

click on ‘For me and my friends’

Give your server a name and a nice picture so you can find it quickly.

You are free to create text channels and categories for different projects and teams.

Go back to the Midjourney server and navigate to one of the “newbies” channels.

Click on the Midjourney Bot and when you see its profile, click on + Add App button.

Then click on “Add to Server” and select your newly created server to add the bot. Make sure to authorize all required permissions, otherwise the prompt codes won’t work.

Oh, and don’t forget to indicate that you’re human…. or not 😉

Important information about payments via Discord (and via the website)

https://www.midjourney.com/account biedt verschillende abonnementsopties aan:

1. Basic Plan – $10/month (€9.20/month)

3.3 GPU hours per month
- Approximately 200 images in standard quality
- Relaxed mode available
- Ideal for hobbyists

2. Standard Plan – $30/month (€27.60/month)

15 GPU hours per month
- Approximately 900 images in standard quality
- Fast mode available
- Perfect for regular users

3. Pro Plan – $60/month (€55.20/month)

30 GPU hours per month
- Approximately 1800 images in standard quality
- Turbo mode available
- Stealth mode (private generation)
- For professional users

As a new user, you get 25 free prompts to explore the platform. After that, you need a subscription to continue working.

The main differences at a glance:

The biggest difference between the subscriptions lies in the number of “Fast GPU” hours. In “Fast” mode, images are generated faster than in “Relax” mode. Important to note: in the Basic subscription, you don’t have access to “Relax” mode. You get 3.3 Fast mode hours in the Basic Plan, 15 hours in the Standard Plan, and 30 hours in the Pro Plan.

Once your Fast hours are used up, you have two options: you can purchase additional hours for €4 per hour or switch to Relax mode. Relax mode has unlimited hours! From personal experience, I can say that Relax mode is sufficient for me, and therefore I don’t need to purchase additional hours.

MIDJOURNEY

Before we dive deeper into this Midjourney guide from Immersive Lab at AP University College, let’s first take a moment to understand how AI-generated images work and how they are created. We’ll examine the advantages and disadvantages of this technology and look at the rules that apply.

In a nutshell:
AI image generation is a technique in which artificial intelligence algorithms are used to create new images. The process relies on a large set of existing images to train a model, which can then generate new visuals based on learned patterns, models, and structures.

⚠️ The process of AI image generation consists of three major steps:

Step 1: Data Collection
First, data is collected. A large dataset of existing images is gathered to train the AI model (for example, to be used in ComfyUI AI / Stable Diffusion). This dataset can include photographs, paintings, and digital illustrations—depending on the intended result.

Step 2: Model Training
The collected visual data is used to train an AI model, usually a deep neural network. These networks are capable of recognizing complex patterns and structures in the data.
This can optionally be done using tools like Kohya.ss.

During training, the AI model learns to generate images in what is known as latent space—a compressed and abstract representation of the images. By manipulating and adjusting points within this space, the model can generate new images.

Once the model is trained, the AI system can generate new images based on your personally trained model. This is done by selecting random points from latent space and decoding them into visually meaningful results. This process can be carried out via Stable Diffusion and ComfyUI AI.

Step 3: Post-Processing
Various post-processing techniques can be applied to the generated images—such as adjusting colors, enhancing details, or adding artistic filters to achieve the desired visual effect.

⚠️ Note: You can’t train custom models directly in Midjourney yet, but you can work with consistent characters, scenarios, and styles. More on that later in this guide.

To make this process easier to understand, we can break it down into three basic elements or concepts:

Numbers and Pixels
Computers operate using numbers (as you may know: 010011001100…). Abstract concepts like text or images must be represented numerically. Each pixel in an image holds color information represented by three numbers: Red, Green, and Blue (RGB).
For the techies, check out color encoding and the RGB color model for more depth.
Image Grids
Images are composed of grids of pixels, and the color of each pixel is determined by specific combinations of RGB values. By altering these values, we can change colors or create new shapes.
Noise
Noise refers to random pixel color values that create blur or distortion. Adding noise involves assigning random values to pixels; removing noise restores cohesion using learned patterns.

In tools like Stable Diffusion, what happens behind the scenes is this:
A text encoder interprets your prompt and identifies important concepts. Then, the image generator—guided by the encoder—uses a diffusion process to produce the final image. Midjourney follows a similar approach.

AI image generators have revolutionized the creative industry through their wide range of applications and, above all, their accessibility. These tools offer benefits for both companies and individuals, although their effectiveness depends on the specific project and its goals. Not every benefit applies to every project.

Let’s take a look at the main advantages of AI image generation:

Time-saving & Volume – AI can generate images in minutes, massively accelerating the creative process.
Cost-effective – Reduces the need for human designers or artists, resulting in significant cost savings—especially beneficial for small businesses or startups (although this can backfire due to the addictive nature of the “rabbit hole”).
Boost for Creative Blocks – AI generators can produce surprising, unique visuals, inspiring fresh ideas that humans might not have considered.
Accessibility – They make the creative process more approachable for people without design or artistic skills.

A breathtaking surreal landscape that represents the essence of MidJourney AI – a fusion of art, technology, and imagination. A dreamlike world where colors flow seamlessly like digital brushstrokes, forming ethereal shapes and futuristic AI-generated sculptures. In the sky, a glowing neural network pulses with creativity, connecting abstract ideas and artistic inspirations. Floating islands showcase different art styles, from cyberpunk to Renaissance, each representing the vast possibilities of AI-generated art. The scene is illuminated with a soft yet futuristic neon glow, evoking a sense of endless exploration and limitless creativity. A subtle, elegant text overlay reads ‘MidJourney Guide’ in a sleek, futuristic typeface, blending harmoniously with the design. –v 5.2 --ar 16:9 --q 2

Midjourney is essentially an AI art generator. The platform uses machine learning to transform text descriptions into visual imagery. Users describe their ideas via an interactive Discord bot, and the AI turns them into unique artworks. Recently, frequent users also gained access to a beta web version.

How Does It Work?

The process is simple:
Users type a description (prompt) of what they want to see, and Midjourney’s algorithms analyze the text to identify visual elements. These are then combined into a completely new, personalized image.
(Command: /imagine prompt: ...)

Why Choose Midjourney Now?

More Artistic Flair – Unlike DALL·E 2, which leans more toward photorealism, Midjourney excels at producing artistic interpretations and styles. With recent updates and the new RAW mode, it is increasingly competing with DALL·E.
Extensive Possibilities – With over 800 supported styles and 700 artists in its database, Midjourney offers unmatched creative freedom. These styles are pre-trained, which raises some ethical concerns we’ll discuss later in this guide.
User-Friendly – Its simple interface and wealth of tutorials make it accessible to both beginners and advanced users. This includes official Midjourney guides as well as tutorials from many YouTubers.
Versatility – Suitable for everything from hobby projects to professional production work.

A major strength of Midjourney is the rapid pace of development. New features are frequently added, and the AI model itself is continually improving, enhancing image quality over time.

💡 Pro tip: Midjourney works via Discord, which means you become part of an active community where you can get inspiration and learn from others. This is one of the reasons Midjourney is such an easy-to-access tool.

One important thing to keep in mind is that—despite its image generation strengths—Midjourney is not optimized for rendering text within images. This is due to several factors:

Artistic Focus – Midjourney’s algorithm is designed for visuals, so text often appears stylized or distorted.
Legibility – Text in Midjourney-generated images is often hard to read or incomplete.
Inconsistency – Letters and words may blur or warp, especially in longer sentences.

For projects where text is crucial, consider:

Adding text afterwards using tools like Photoshop
Focusing on visuals first and treating text as a separate design element

⚠️ Tip: Keep text use in Midjourney prompts minimal for the best results. For more on generating optimal text prompts, we’ll dive into the Ultimate Prompt later in this guide.

Midjourney

Before we fully dive into Midjourney, let’s first briefly discuss the different versions currently available.
As I mentioned earlier, Midjourney evolves almost as fast as Adobe, so keeping track of all the developments week by week is certainly a challenge.

First of all, you can still use all previous versions of Midjourney. These are easy to select within the settings on Discord—we’ll come back to that later in this guide.
In this section, I’ll show you how quickly Midjourney has developed and how the different versions compare to one another. You’ll also see just how dramatic the difference is from version 1 to version 4.

To make the comparison as clear as possible, I’ll use the same prompt each time:

A medieval castle on a hilltop, surrounded by a lush forest, golden sunset --v {1,2,3,4,5,5.1,5.2,niji 4,niji 5,niji 6,6,6.1}

💡 Pro tip: By using {} (also called permutations) in your prompt, you can generate multiple variations at once in Discord. This saves you from having to retype the same prompt over and over. With a free account, you can create up to 4 variations at a time. If you have a Standard subscription, you can generate up to 10 variations. And with a Pro or Mega subscription, you can generate up to 40 variations in a single run!

Niji models are a special series within Midjourney, developed in collaboration with Spellbrush.
They are entirely focused on Eastern and anime-style art, with an emphasis on an illustrative look. Perfect for creatives who want to generate vibrant, detailed, anime-inspired images.
These models even have their own website and Discord server.

The latest Niji 6, released on June 7, 2024, introduces a host of exciting improvements. It now performs much better with Japanese text—especially short kana characters and simplified Chinese symbols. The images have also become much more detailed, particularly when it comes to typical anime features like eye structure.
And those little quirks or glitches you might’ve seen in earlier versions? They’re now a thing of the past!
— Source: Midjourney Docs – Version

In this section, we’ll guide you through creating your first prompt, upscaling images, generating variations, and saving your first creation.

Whether you’re using Midjourney’s new web app or Discord, you’ll find the same functionality—only the user interface and experience differ.

This chapter covers the essential commands and functionalities of Midjourney. We explain all technical aspects needed for writing effective prompts and using available commands. Since both Discord and web versions share identical functionality, this guide applies to both platforms.

The chapter provides a complete guide for both basic and advanced functionalities, with practical examples and technical specifications for optimal platform use.

Basic Commando’s

The essential commands for using Midjourney to generate your first images:

/imagine – The primary command for generating images based on textual descriptions. Syntax: /imagine [description]

Lets keep it simple for now Example: /imagine a majestic orange cat, sitting in warm sunlight

Variation Commands

Options for refining generated images:

U1-U4 selections – Increase the resolution of a selected variant
V1-V4 selections – Select from four different interpretations of the given prompt
🔄 option – Generate alternative variations of the selected result

Let’s create a new variation by clicking V1 and transform the warm sunlight into a cool variant. This is called a ‘remix prompt’.

As you can see, the AI maintains the cat and about 80% of its pose. It attempts to apply the new prompt by incorporating ‘cool moonlight’ into the scene. While the image now has more of an evening atmosphere, the cat still appears to be sitting under artificial lighting. We can refine this further by adding additional parameters to our prompt.

Parameters

Now we’ll explore how to add more dimension to our image. We’ll experiment with different parameters to adjust the size, introduce more ‘chaos’, add variation, and make it more artistic – all by simply tweaking various parameters.

For consistency in this guide, we’ll continue using our cat prompt as an example.

1. Aspect ratio

The aspect ratio determines the proportion between width and height of an image. It’s written as two numbers with a colon between them, like 1:1 or 4:3.

By default, Midjourney creates square images, which means they have a 1:1 aspect ratio – equal width and height.
To choose the right aspect ratio, think about whether you want a wide (landscape) or tall (portrait) format. For landscape images, use a larger first number; for portrait images, use a larger second number. — https://docs.midjourney.com/hc/en-us/articles/31894244298125-Aspect-Ratio

a majestic orange cat, sitting in warm sunlight --ar {1:1,1:2,2:1,2:3,3:2,4:3,3:4,4:5,5:4,4:7,7:4,5:8,8:5,9:16,16:9,9:19,19:9,9:20,20:9}

As you can see, Midjourney can create many different aspect ratios. You can actually create even more extraordinary variants, such as true banner formats. Here is an example of 5:1 (which I used to shape the layout of this guide).

a majestic orange cat, sitting in warm sunlight --ar 5:1

2. Chaos

The chaos parameter in Midjourney (called Variety in the web interface) determines how creative the AI gets with your prompts. When you set chaos to 0, which is the default setting, Midjourney will create images that closely follow your prompt instructions. It’s like following a cooking recipe exactly as written – you’ll get predictable, consistent results. One of the fun parameters we use.

However, when you increase the chaos value (anywhere from 1 to 100), you’re giving Midjourney permission to be more creative. Think of it as telling a chef to experiment with a recipe. The higher the chaos value, the more creative freedom the AI takes, potentially leading to unexpected and unique interpretations of your prompt. For example, a prompt for “a red house with a garden” at chaos 0 will give you a traditional red house, but at chaos 100, you might get a surreal architectural masterpiece with nature elements woven throughout the structure.
— https://docs.midjourney.com/hc/en-us/articles/32099348346765-Chaos-Variety

a majestic orange cat, sitting in warm sunlight --c {0,25,50,75,100}

3. No

No means no…

The –no parameter helps you exclude unwanted elements from your generated image. Adding this parameter tells the AI what to avoid including, giving you better control over the output. You can exclude specific objects, colors, textures, or any other elements that don’t match your vision. I will expand the cat prompt with a painting variant and flowers to achieve a certain effect, and then remove something I don’t want. — https://docs.midjourney.com/hc/en-us/articles/32173351982093-No

The '--no' parameter isn't always perfect. Sometimes Midjourney might still include elements you wanted to exclude from your cat image. This can happen due to:

Complexity of the Prompt: The more complex your prompt about the subject the higher the chance that Midjourney might overlook the ‘–no’ parameter.
Algorithmic Limitations: Midjourney’s algorithms are trained on vast datasets, which may sometimes lead to unintended additions in your photo.
If you notice the ‘–no’ parameter isn’t working as expected, here are some strategies:
1. Be More Specific: Instead of just using --no collar, try --no collar, pet accessories, neck wear. (to stay in the cat analogy)
2. Use Negative Weights: You can assign negative weights to unwanted elements for more control. For example, --no dog -5. (This significantly improves results in my opinion)

4. Quality

The Quality parameter (–q) in Midjourney is one of the most crucial settings you can use to optimize your images. This parameter ensures an important balance between image quality and generation speed.

When working with the quality parameter, you have several options available. The lowest setting is 0.25, which provides the fastest generation time but at the expense of image detail. This setting is perfect when you want to quickly test different ideas or sketch concepts.

The middle option of 0.5 offers a good balance between speed and quality. This setting is ideal for daily use and when you don’t necessarily need the highest quality but still want clean results.

The default setting of 1.0 is what most users choose for their projects. This setting produces detailed images without extremely long wait times. It’s an excellent choice for most applications. —https://docs.midjourney.com/hc/en-us/articles/32176522101773-Quality

a majestic orange cat, sitting in warm sunlight --q {0.25, 0.5, 1}

For the most demanding projects, there’s the 2.0 setting. This highest quality setting ensures maximum detail and sharpness in your images. Keep in mind that this setting also takes the most time to generate.

It’s important to know that the quality setting only affects the technical aspects of your image, such as detail and sharpness. It doesn’t change the AI system’s creative interpretation of your prompt.

For the most efficient use, we recommend starting with lower quality settings while experimenting. Once you’re satisfied with the concept, you can switch to a higher quality setting for the final result.

Despite the benefits of higher quality settings, there are some important downsides to consider:

Longer Generation Times: Higher quality settings (especially 2.0) can significantly increase the time needed to generate images

Higher GPU Usage: Quality settings above 1.0 consume more computational resources, which affects your subscription quota faster !

Cost Efficiency: Using maximum quality for initial concepts or experiments can be wasteful of subscription credits

Therefore, it’s crucial to choose the appropriate quality setting based on your specific needs and project phase.

a majestic orange cat, sitting in warm sunlight --q {0.25, 0.5, 1}

5. Stylization

The Stylization parameter (–s) in Midjourney allows you to control the artistic style of your generated images by adjusting how strongly Midjourney’s default aesthetic style appears. You can set values ranging from 0 to 1000, with 100 being the default setting that most users start with…

— Midjourney Stylize

Based on our experience, this is one of the most fun settings to experiment with next to the quality parameter, as it allows you to dramatically transform the artistic style of your images from photorealistic to highly stylized interpretations.

Tips for Effective Use

Understanding the different stylization levels is crucial for achieving your desired results. The lowest level, zero, produces highly realistic images that are ideal for photographic results. Moving up to level 100 provides an excellent balance between realism and artistic interpretation. The default setting of 250 introduces a more distinctive artistic style, while setting the parameter to 1000 creates the most artistic and stylized interpretations possible.

Despite the advantages of different stylization settings, there are several important considerations:

Realism vs. Artistic: Higher stylization settings (especially 1000) can significantly reduce realistic details
Consistency: Different stylization levels can result in very different interpretations of the same prompt
Desired Effect: Too high or too low stylization can affect the intended result of your project
Experimentation: It may be necessary to test different stylization levels to find the right balance

Therefore, it is crucial to choose the right stylization setting based on your specific needs and project goals.

In this second section, we will explore some technical parameters that may not seem immediately creative, but are nonetheless fascinating to examine. We’ll dive into powerful features like seeds for reproducibility, repeaters for pattern creation, and tiling for seamless textures.

1. Tile

The –tile parameter in Midjourney creates seamless, repeatable patterns where edges match perfectly when tiled side by side. This feature is valuable for designers creating continuous patterns for textiles, backgrounds, and wallpapers.

This feature works best with illustrations rather than photographs. As you can see in the example below using our cat prompt from the first chapter, the tiling effect is not effective when applied to photographic images.

a majestic orange cat, sitting in warm sunlight --tile

Let’s try this with an illustration instead. I’ll add the word “illustration” to the prompt to let Midjourney know directly that we want an illustrative style.

an illustration of a majestic orange cat, sitting in warm sunlight --tile

Still not quite what we want. It may take a few attempts to achieve the desired result.

an illustration pattern of a majestic orange cat, sitting in warm sunlight --tile

We picked one of these and turned it into a pattern. Now we have what could be a perfect wallpaper for a crazy cat lady 😉

Let’s try again, without the sun.

an illustration pattern of a majestic orange cat --tile

We picked one of these again and turned it into a pattern.

Much better 😉

Want to check if your pattern tiles seamlessly? It’s super easy! You can upload your pattern to the Checker tool – a handy, free website where you can see in just a few clicks if everything lines up perfectly. This way, you’ll know for sure that your pattern will look flawless whether you use it for wallpaper, fabric, or whatever else you have in mind!

2. Seed

We’ll dive into powerful features like the seed parameter for reproducibility – which allows you to regenerate exactly the same image by using the same seed number.

By adding a specific seed number to your prompt (for example –seed 123456), you can consistently use the same base for your image generation. This is particularly useful when you want to make small adjustments to an existing generation without starting from scratch. — https://docs.midjourney.com/hc/en-us/articles/32604356340877-Seeds

Experimenting & Testing: Lock a seed when testing different elements in your prompts. It’s like having a control in an experiment. This makes it easier to see how changes in your prompt affect the result.

Not Always Predictable: Seeds may behave unexpectedly, so they shouldn’t be relied on for the same results over different prompting sessions. Assume that if you take a break and come back later, your seed may no longer work as expected.

a majestic orange cat, sitting in warm sunlight --seed 100

Now the client finds the position of the cat perfect and the way the sun shines. They want to adjust this, but the cat needs to become a dog. If we start completely from scratch with a random seed, we’ll get completely different images, but we’ll use the seed from the cat, which is 100, and let’s see what happens.

a majestic dog, sitting in warm sunlight --seed 100

The viewing direction of the dog is almost identical, as are the color and sun intensity. So we’re definitely heading in the right direction.

To find the seed number of your generated image, click the ‘Add Reaction’ button below the image and select the envelope emoji (✉️). If you can’t find the emoji right away, you can use the search bar at the top of the emoji window.

The Midjourney Bot will then send you a Direct Message with the seed number. If you do not receive a direct message, make sure to check your Privacy Settings for the Midjourney Discord server and confirm that “Direct Messages” is toggled on to allow the Midjourney Bot to send you a direct message

3. Weird

The weird parameter offers an exciting way to add unexpected and artistic elements to your images. This unique feature allows you to control how unconventional your generated images become. These work better with illustrations or art forms in our experience.

By adjusting the weird parameter in your prompt (using –weird followed by a value from 0 to 3000), you can explore different levels of surrealism. Lower values produce more conventional results, while higher values create increasingly abstract and unexpected interpretations. For example, –weird 100 might add subtle artistic twists, while –weird 3000 could transform your prompt into something completely surreal.

Keep in mind that weird is an experimental feature, and what's "weird" may change over time and Versions.

an illustration of a majestic orange cat --weird {0,50,100,500,1500,3000}