Midjourney: How to Make AI-Generated Art

Have you ever wished you could transplant your imagination into vivid visuals? With the advancements in AI image generators, that dream is now a reality. And if you haven’t met Midjourney yet, you probably should. Midjourney is one of–if not THE–leading AI art & image generation tools out there, and it allows you to bring your ideas to life with simple text descriptions known as “prompts”.

So how does it work? Let’s explore the features and potential of Midjourney, and how it’s become a radical new tool for visual creativity.

Pink Horn… I wonder where she gets her name…

Meet Midjourney

Midjourney stands at the forefront of AI image generators, using the power of a large language model (LLM) trained on an extensive dataset of images and text. What does that mean? It means the machines learned–not copied, there is a distinction–from the data that is fed to it. Just as any student would, only in machine capacity, which is unimaginable in human terms. Hundreds of millions even billions of data sets and sources are used to train the AI so that its wealth of knowledge and understanding can be as vast as it is. There is some controversy around this, but that is not the point of this blog.

By combining natural language processing with cutting-edge image generation techniques, Midjourney can produce images in various styles, ranging from realistic to abstract, illustrative to surreal. And beyond. Although it continues to evolve with each new iteration bringing new innovation to the platform, this remarkable tool has already wowed its millions of users with its ability to generate just about anything you can come up with. (There’s that training result.)

Cyberpunk Noir Detective, Valentine

Remember, Midjourney (and art generators in general) can generate images, because they have learned how. They are not copying, they are not patchworking together pieces of existing artworks, they have learned how to produce their own images, just as a novice painter learns how to render landscapes in oil by following along with Bob Ross on the Joy of Painting.

The AI is the painter, the datasets are Bob Ross.

Wow, I went there.

Because of this, Midjourney can produce visual outputs in as many styles as you can think of. With the ability to generate images from object descriptions, style instructions, references to detail, color, camera models and settings, and a vast wealth of other prompting parameters, Midjourney opens up endless possibilities for artistic exploration and expression.

Endless.

So how does it work?

Midjourney offers incredible potential for creating realistic and stylistic images.

Midjourney: It’s All About the Prompting

AI generators like Midjourney work on a basic text-to-image premise. Users”prompt” with text describing subjects, scenes, people, or even instructions (not a narrative) on how to style and compose an image. This versatility allows artists, designers, and enthusiasts to explore a wide range of creative ideas.

The ease of producing an image through prompting honestly depends on one’s purpose and expectations. If your aim is to just generate images with no clear direction, Midjourney will be a breeze and super impressive. But for creatives who are working on specific projects–for instance, consistent generations for an AI film, or panels for a graphic novel, or character portraits, interior designs, the re-creation of World War II photos–anything with a more meticulous concept–prompting in Midjourney becomes an art unto itself. (We call it ‘promptography’ or ‘prompt engineering.’)

There is a lot to learn where fine-tuned prompting is concerned. It’s easy to make it sound like you input text, and -boom- AI gives you the perfect image. Maybe in the beginning it feels that way, because that is the general capability, but the more involved you get with an AI generator like Midjourney as a tool for creation, the more nuance and functionality you must learn, discover and practice. Like any other craft.

We have a whole growing section of articles that explores the various features, tactics and techniques of using Midjourney. It’s written and curated by our Pink Horn team, which includes traditional artists and photographers who have embraced these new AI tools to further their art and creative process. It’s a growing section, and probably our favorite here at Pink Horn because there is so much still to learn and discover and experiment with. And that’s exciting creatively.

Midjourney Promptography 101: Mind the Prompts: Hyper realistic, photography shot through an outdoor window of a rainy dystopian city apartment with pink and blue neon sign lighting, window glares and reflections, depth of field, a cyberpunk girl with pink hair, sitting at a table, portrait, Kodak portrait 800

How to Get Started with Midjourney

Getting started with Midjourney is a straightforward process – this is the most basic of the basic ‘how to’:

• Create an account on the Midjourney Discord platform (see below on pricing options).
• Once logged in, access the prompt bar in one of the #generate threads or through the Midjourney Bot and use the /imagine command.
• Type in your desired text prompt, detailing your envisioned image.
• Midjourney will generate a selection of images based on your prompt.
• Select the generated images that resonate with your creative vision.
• Upscale, remix, and vary to your liking.

Prompt: cyberpunk fashion model, alternative style, tattoos. Neon elements. High detail. –ar 4:5

Midjourney Basics: Modes

Midjourney offers various settings and modes that users can adjust to customize their image generation experience. These settings and modes provide users with flexibility in tailoring their image generation process to their preferences and requirements. Users can choose the version, image quality, style, privacy, and generation speed settings that best suit their creative needs.

Here’s a summary of the key Midjourney modes and settings:

• Version Settings: Users can switch between different versions of Midjourney (MJ Version 1-5.2). Each version may offer different capabilities or styles. After switching, all generated images will use the selected version by default.

• Image Quality Settings: This setting affects the details of the generated image but not its resolution. Users can choose from different image quality settings, such as “Base quality.” Default setting is recommended for most users.

• Style Settings: Determines how closely the generated image matches the prompt. Lower values align more with the prompt’s description, while higher values increase artistic quality. Users can adjust this setting to achieve the desired balance between prompt alignment and artistic quality.

• Privacy Settings: By default, set to “Public.” Pro users can set it to “Stealth Mode,” which may limit the visibility of their generated images. Basic and Standard users cannot use “Stealth Mode.”

• Remix Settings: Remix Mode is an experimental feature. Users can click the “Variation” button in Remix Mode and provide a prompt. The generated image retains the structure of the original image but changes the subject based on the prompt. This mode allows users to create variations of existing images.

• Generation Speed Settings: Default setting is “Fast Mode.” Standard and Pro users can switch to “Relax Mode.” In “Relax Mode,” image generation may take longer but is unlimited for Standard and Pro users. Basic users have “Fast Mode” as the only option.

Learn how to prompt to get results.

Tips for Enhancing Your Midjourney Results

But let’s go deeper. There are many techniques to enhance your results, so let’s begin there:

• Compose Concise and Thoughtfully Structured Prompts: When using Midjourney, the order and brevity of your prompt matter. The system places more significance on the earlier parts of your description. If there’s a specific element you want to include in the image, make sure to place it near the beginning or middle of the prompt. Concise prompts tend to yield better results, so focus on being clear and to the point.

• Use Lighting and Lens Descriptions in Your Prompts: To achieve photorealistic results, mastering lighting and lens descriptions is crucial. Midjourney understands prompts that mention specific cameras, lenses, focal lengths, and even film stocks. Phrases like “high key,” “low key,” “high contrast,” and “low contrast” can influence the outcome. Powerful keywords include “global illumination,” “natural light,” and “uplight,” which can create specific lighting effects. For example, using “uplight” can give a softer look to the image.

• Give Midjourney Style Suggestions: One of the most effective ways to enhance your results is by specifying a style for Midjourney to emulate. Phrases like “film still” and “in the style of” or “style by” allow you to reference TV programs, movies, or specific styles. Adding a style description can significantly improve the outcome. Experiment with different styles and combinations to achieve the desired effect.

• Upload a Reference Image to Midjourney: When attempting to generate complex images or interactions, Midjourney may require a reference image to produce the desired result. You can provide a reference image by uploading it to the prompt box within Discord. This image will help guide Midjourney in generating a closer representation of your intention. Be patient and iterate, as it may take multiple attempts to achieve the desired outcome.

• Try, Try Again: Achieving excellent results often requires patience and experimentation. Don’t hesitate to regenerate images or create variations based on the ones you like. Use the “–prefer remix” command to generate variations with different seed values or additional instructions. This feature offers endless possibilities for manipulating existing images and refining your results.

Valentine After the Dystopian Wars: hyper realistic cyberpunk Tom Hardy with a fade undercut, dystopian city, reflecting the city lights that illuminate the wet pavement, portrait style, Nikon D6 high shutter speed action shot, neon pink and blue lighting

Using Parameters in Your Prompts

Midjourney offers various parameters that users can use to customize their image generation prompts effectively. These parameters allow you to customize your image generation, from specifying aspect ratio to controlling randomness, stylization, and even switching to the specialized Niji model for anime-style images. With these parameters you can fine-tune your prompts, ensuring they get the desired style, content, and consistency in the generated images. Each parameter serves a specific purpose, allowing users to create a wide range of image variations with Midjourney.

Here’s a summary of the key Midjourney parameters:

Aspect Ratios Parameter (--ar):

• How It Works: This parameter allows you to specify the aspect ratio of the image you want to generate. Aspect ratios affect the shape and composition of the generated image.
• Usage: Include --ar in your prompt followed by the desired aspect ratio. For example:

  cyberpunk girl with pink hair --ar 5:4

No Parameter (--no):

• How It Works: This parameter excludes specific elements from the generated image. For instance, if you don’t want trees in the image, you can input --no trees in your prompt.
• Usage: Add --no followed by the element you want to exclude. Example:

  cyberpunk city street --no people

Chaos Parameter (--c):

• How It Works: Chaos controls the randomness of the model. A higher value (e.g., --c 100) produces more unexpected results, while a lower value (e.g., --c 10) increases consistency.
• Usage: Add --c followed by a number from 0 to 100 to your prompt. Example:

  cyberpunk human cyborg hybrid --c 100

cyberpunk human cyborg hybrid –c 10 (left), cyberpunk human cyborg hybrid –c 100 (right)

Stylize Parameter (--s):

• How It Works: Stylize controls the degree of stylization in the generated images. A lower value (e.g., --s 50) aligns more with the prompt’s description, while a higher value (e.g., --s 100) increases artistic elements.
• Usage: Include --s followed by a number in your prompt to control stylization. Example:

  colorful risograph of a cyberpunk robot --s 50

Niji Parameter (--niji):

• How It Works: The Niji parameter is used to activate the Niji model, which specializes in generating anime-style images. Add --niji at the end of your text prompt to use the Niji model.
• Usage: Simply add --niji at the end of your text prompt to enable the Niji model. Example:

  cyberpunk rock star --niji

cyberpunk rock star –niji (left) cyberpunk rock star Johnny Silverhand –niji (right)

Notable Features:

As of writing this, the latest version of Midjourney is V5.3, and here is a breakout of the most notable features and functions:

1. Improved Image Quality: MidJourney V5 offers much higher image quality compared to previous versions. The default resolution is 1024 x 1024 pixels, which is double the resolution of V4, resulting in sharper and more detailed images.
2. Realism: V5 is capable of generating extremely realistic images. It pays close attention to details described in prompts, making the generated images appear more lifelike. Users have greater control over the final image’s realism.
3. Stylistic Range: MidJourney V5 retains the artistic characteristics seen in V4 but also allows users to fine-tune the stylistic elements of generated images. By adjusting the “stylize” parameter, users can control the level of stylization in the images.
4. Remix Feature: The “Remix” feature, which combines two uploaded images, has been improved in V5, resulting in more impressive and cohesive image combinations.
5. Aspect Ratio Options: V5 offers unlimited aspect ratio options, allowing users to create images with unique and extreme aspect ratios, from 30:1 to 1:50.
6. Image Weight Control: Users can now control how much an image reference influences the final image by using the “image weight” parameter. This provides more control over the blending of an image reference with the generated image.
7. Tile Parameter: The tile parameter has been reintroduced in V5, allowing users to create seamless patterns and textures for various purposes, such as wallpapers and gaming textures.
8. /describe for Image Descriptions: Midjourney introduced a new feature called /describe for image-to-text. Users can use the /describe command and upload an image to receive four text prompts describing the image, which they can then employ in their own prompts in language Midjourney understands.
9. Repeat and Permutations Features: Midjourney added two new advanced prompting features: Repeat and Permutations. These features generate multiple jobs and are available to Pro Members and during fast hours.
10. Midjourney’s Vary Region: is a new feature that allows users to alter selected parts of their generated images using AI. It can be accessed through Midjourney’s Discord Bot or by enabling the Remix mode. Users can modify prompts to refine specific regions of the image, introducing flexibility and enabling improvements to desired areas.

Formidable Netrunner, GhostCipher: hyperrealistic, high-definition portrait of black cyberpunk woman, cyberdreads and bright ice-blue eyes with a superior, intense gaze. Skin texture, and undertones. She is in a sleek, cyberpunk club, surrounded by neon elements. There is power in her face. emulate Nikon D6 high shutter speed action shot, dramatic neon lighting

Midjourney Inpainting

Midjourney’s inPainting feature is a powerful addition, offering users the ability to fine-tune and edit their images with ease. It’s a valuable tool for both creative expression and practical image enhancement.Here’s a summary of its key capabilities:

1. Replacing an Element: Users can replace elements within an image by modifying the prompt in a designated box. This feature offers a seamless transition from one element to another, allowing for creative image composition.
2. Adding an Element: InPainting enables users to introduce new elements into an image, seamlessly blending them into the composition. This is useful for scenarios like adding people to landscapes to convey a sense of scale.
3. Removing an Element: Unwanted elements in an image can be easily eliminated by highlighting the specific area to be removed and excluding it from the prompt.
4. Repairing an Element: InPainting also allows for the correction of imperfections within an image. Users can select a problematic area and effectively redraw and repair it. This is particularly valuable when precision is required.

Pro Tips for Optimal Results:

• InPainting works best on large regions of the image (20% to 50% of the image).
• For subtle changes, consider using “Vary(Subtle)” instead of drastic alterations.
• Changing the prompt is most effective when it matches the image context.
• InPainting can be used as a general-purpose editing tool.

Midjourney is Ever-Evolving Its Features

Midjourney’s Zoom Out

Midjourney’s Zoom Out feature empowers users to create expansive AI-generated scenes based on text prompts, providing various zooming and aspect ratio options for customization. It’s a powerful tool for unleashing creativity in AI-assisted content creation.

1. Expanding AI-Generated Content: With Zoom Out, Midjourney enables users to expand the AI-generated content they’ve created. It’s similar in concept to zooming out from a focal point in a camera shot, revealing a broader scene filled with AI-generated elements.
2. Text-Based Input: Unlike traditional image editing, where you start with an image, Midjourney’s approach begins with a text prompt. Users describe their image concept to the AI, and it brings that vision to life.
3. Zooming Options: Once your image is generated, you can use the zooming function. Midjourney offers preset zoom options like 1x, 1.5x, and 2x. These controls allow you to stretch the AI’s creative capabilities.
4. Aspect Ratio Modification: The “MakeSquare” command allows users to adjust the aspect ratio of the image, transforming it into a perfect square or extending it vertically or horizontally as needed.
5. Custom Zoom: For more control, the “Custom Zoom” button lets you specify the degree of zooming, accepting values between 1-2. It also allows you to modify the aspect ratio using the ‘–ar’ parameter without zooming out. You can even tweak the initial prompt for unique results.

Rebellion Leader, Frontier: cyberpunk Idris Elba with cyberdreads, powerful eyes with a hard, dark, intense gaze, tattoos, dystopian fashion, neon signs, reflections, defiant, high-definition, realism

Pros and Cons of Using Midjourney

When considering the use of Midjourney, an AI art tool that generates images from text descriptions, it’s essential to weigh the advantages and disadvantages. Here are the pros and cons to help you make an informed decision:

Pros:

• Creativity: Midjourney provides a platform for artists, designers, and enthusiasts to unlock their creative potential. It transforms text prompts into visually captivating images, enabling users to explore and express their ideas with ease.
• Versatile Image Generation: With the ability to produce images in various styles, from realistic to abstract and surreal, Midjourney offers a broad range of creative possibilities. It caters to different artistic preferences, allowing users to experiment and push the boundaries of their imagination.
• Feedback-Driven Improvement: One of Midjourney’s key strengths is its feedback-driven image generation. Users can rate images and provide specific feedback, actively participating in refining the artistic output. This iterative process helps Midjourney learn and adapt, ensuring a more personalized and satisfying user experience over time. There is also an active Midjourney community on Discord to discuss all manner of features and functions and provide support.
• Ease of Use: Midjourney offers a user-friendly interface, making it accessible to both beginners and experienced artists. The prompt bar simplifies the image generation process, and the ability to save favorite images to a personal gallery enhances convenience.

Cons:

• Dependence on Text Prompts: Midjourney’s image generation relies heavily on text prompts. While this provides flexibility, users need to invest time and effort in crafting clear and descriptive prompts to achieve desired results.
• Controversy: Although Midjourney can produce stunning images, it may lack the nuanced touch and artistic intuition that a human artist can bring to their work. The generated images may not always capture the full depth of human creativity and expression.
• Subscription Cost: Midjourney is not free to use. Users need to subscribe to a monthly plan to access its full features. It offers four subscription plans with different features and pricing options, the cost may be a consideration for some users.

Covert Operative, Jinx: cyberpunk fashion model, long bright yellow hair, alternative style, tattoos. Neon elements. High detail. –ar 4:5

Midjourney 6 is Coming

Midjourney 6 is the next generation set to bring significant enhancements and capabilities. Here’s a summary of what to expect with Midjourney 6:

• Higher Image Resolution: Midjourney 6 will offer a maximum image resolution of 2048×2048 pixels, nearly double the resolution of the previous version. This means you can create more detailed and higher-quality images.
• Smarter Natural Language Processing: The AI in Midjourney 6 will have a refined understanding of text prompts, enabling it to generate more precise and realistic images based on your descriptions.
• More Control Over Variations: Users can expect increased customization options for image variants. This may include the ability to change colors, styles, or angles of generated images, giving you greater creative control.
• Improved Hand Generation: Midjourney 6 will feature improved hand generation, making it easier to create images with realistic hands, enhancing the overall realism of your creations.
• 3D Models: One exciting addition is the ability to generate 3D models. This opens up new creative possibilities for users, allowing them to work with three-dimensional art and designs.
• Video Generation: Midjourney 6 might introduce the capability to generate videos from text inputs. This feature has the potential to revolutionize storytelling, filmmaking, and education, enabling users to create video content directly from their textual descriptions.
• Speed and Reliability: Midjourney 6 will offer improved speed and reliability compared to previous versions. Users can expect faster image generation, reducing wait times and allowing for more efficient creativity.
• User-Friendly Interface: Midjourney 6 will also feature a redesigned website with a more user-friendly and efficient interface, enhancing the overall user experience.

Antagonist, VoidReaper: hyperrealistic, high-definition portrait of cyberpunk guy with vibrant blue hair and devious bright blue eyes with a hard, dark, intense gaze. Skin texture, water droplets, dirt grit. dystopian city, heavy rain Leica IIIc f/4, 1/200 shutter speed, color film, high-resolution photography, dramatic lighting

Midjourney’s Pricing:

Midjourney no longer offers a free tier or trial period to new users. Previously, it did offer a free trial that allowed users to generate 25 images for free, but this has been discontinued. To use Midjourney, you’ll need to subscribe to one of three paid subscription plans. Each plan offers different features and usage limits.

The plans offer varying GPU allowances and modes for generating images. The higher-tier plans provide more GPU hours and additional features like private generations. It’s important to note that even with the higher-tier plans, there may be monthly limits on the number of images generated in Fast mode. Midjourney also allows users to purchase additional hours of GPU compute time if needed.

Here are the details:

Basic Plan:

• Monthly Cost: $10
• Annual Cost (20% off monthly): $96
• Fast GPU Allowance: 3.3 hours/month
• Relaxed Mode Allowance: 0 minutes
• Private Generations: No
• Images Publicly Visible at All Times: Yes

Standard Plan:

• Monthly Cost: $30
• Annual Cost (20% off monthly): $288
• Fast GPU Allowance: 15 hours/month
• Relaxed Mode Allowance: Unlimited, dynamic queue system
• Private Generations: No
• Images Publicly Visible at All Times: Yes

ProPlan:

• Monthly Cost: $60
• Annual Cost (20% off monthly): $576
• Fast GPU Allowance: 30 hours/month
• Relaxed Mode Allowance: Unlimited, dynamic queue system
• Private Generations: Available via Stealth mode
• Images Publicly Visible at All Times: No

Mega Plan:

• Monthly Cost: $120
• Annual Cost (20% off monthly): $1152
• Fast GPU Allowance: 60 hours/month
• Relaxed Mode Allowance: Unlimited, dynamic queue system
• Private Generations: Available via Stealth mode
• Images Publicly Visible at All Times: No

Midjourney does come at a cost, and users will need to choose a plan that suits their needs and budget.

CyberChonk Prompt: cinematic portrait of a fat orange and white cat in a cyberpunk city, surrounded by neon signs, blue, pink, puddles, and wet reflections. high-definition details –ar 4:5

Yes. Midjourney is Really, Really Impressive

Is generating an AI image easy? It can be. Is generating an AI image hard? It can be.

The truth of it comes down to input/output. Like with anything AI, you get out as good as you put in. There is still an effort, still a human creator, still a learning curve, still a process, still a craft. It’s different from other crafts, it’s its own now, to be learned and understood on its own terms, but where you take it and how you make it work for you, is on you–the human.

Midjourney represents a significant leap forward in AI-generated art. It is really impressive. With every iteration, the platform keeps honing and innovating its functions. With its ability to generate images in more styles than you can shake a stick at, Midjourney caters to a wide range of artistic preferences and allows creatives to explore and experiment with different artistic visions in a dynamic way. And the generated image doesn’t even mean that’s the end result–perhaps it’s just a reference point in a larger artwork. Maybe even a traditional artwork. I know personally, I’d like to paint from many of these image generations I’ve prompted. I see so much potential to incorporate AI as a tool in my creative process now, not to detract, but to add.

As Midjourney continues to evolve, we can only imagine the incredible potential it holds for artists, designers, and enthusiasts. And the truly wild thing is to realize, we’re really just at the beginning of this journey.

More on Midjourney:

• How to Achieve Cinematic Photorealism with Midjourney Prompts
• How to Unlock The Art of Photorealism in Your Midjourney Prompts
• Colors that Speak: How To Elevate Your Midjourney Prompts with Color Theory
• A Closer Look at the New Midjourney inPainting Feature
• How to Use the Midjourney Pan & Blend Features to Render Your Best Self
• It’s ALIVE! How to Use Midjourney Seed for Consistent Imagery
• How to Generate Full-Body Images with Midjourney: 5 Expert Tips