Ultimate MidJourney Beginner’s Guide!

Midjourney is a generative artificial intelligence tool created and hosted by the San Francisco-based independent lab Midjourney, Inc. Midjourney generates images from natural language descriptions, called prompts, in a manner similar to OpenAI's DALL-E and Stable Diffusion.

On July 12, 2022, the tool entered open beta. The Midjourney team is led by David Holz, co-founder of Leap Motion. In August 2022, Holz told The Register that the company was already profitable. Users create art with Midjourney using Discord bot commands.

Currently, Midjourney can only be accessed via a Discord bot on their official server, by sending a direct message to the bot or by inviting it to a third-party server. To generate images, users use the /imagine command, then enter a prompt. The bot then returns a set of four images that users can choose to enlarge. Midjourney is also working on a web interface.

How to use Midjourney?

Midjourney is an artificial intelligence tool that transforms textual descriptions (the “Prompts”) into personalized images on Discord. This part will explain you how to use the Midjourney bot to make the most of its features.

Above all, it is important to respect the code of conduct by being courteous, by avoiding creating offensive images and by showing respect towards the other users and the Midjourney team. You risk being banned quite quickly if you do not respect the code of conduct.

Step 1: Join the Midjourney Discord server To begin, connect to your verified Discord account and join the Midjourney server by clicking on “Join the beta” from the Midjourney.com website.

You must have a Discord account (a kind of Skype that allows discussions in the form of a chat room) to use Midjourney. It's free and quite easy to use, you can download it on Windows, Mac and smartphone here : https://discord.com/download

Here is the Discord interface and the Midjourney space:

Step 2: Discovering the lounges for beginners Once on the official Midjourney server, choose a “newbies-#” lounge to familiarize yourself with the bot's features. Just click on one of the available rooms to have access to it:

You can also invite the Midjourney Bot on your own server, here is how to create your own Midjourney rooms.

Step 3: Use the /imagine command The Midjourney bot is used thanks to commands. The /imagine command allows you to transform a short textual description into a unique image. To use it, type /imagine prompt: followed by the description of the desired image. The bot will then generate 4 images that describe your Prompt.

Explore the different versions of models available at Midjourney

Midjourney regularly offers new versions of templates to improve the efficiency, consistency and quality of its renders. The most recent model is activated by default, but it is possible to use other models by adding the –version or –v parameter, or by using the /settings command and choosing a model version. Each template has its own strengths to handle different types of images.

Latest Model: Midjourney V5.2

The latest version of Midjourney to date was released publicly on 23/06/2023, here are the main changes:

  1. Improved aesthetics and sharper images: Midjourney 5.2 brings improvements to the visual quality of generated images, making them sharper and more aesthetically pleasing.
  2. Slight improvement in text consistency and comprehension: Midjourney's text comprehension capabilities have been enhanced in version 5.2, resulting in greater consistency between text indications and generated images. Prompts can be shortened and simplified in this version.
  3. Greater diversity: In Midjourney 5.2, there is greater diversity in the models generated. Previously, if no gender was specified in the prompt, the template generated mainly female models. With this update, however, there's a mix of male and female models. If you specifically wish to generate models of a particular gender, you can include this information in your indication.
  4. Enhanced Stylized functionality: The “Stylized” command has been updated in Midjourney 5.2 to have a greater effect on the amount of stylization applied to generated images. The “Stylized” command now ranges from 0 to 1000, with 100 being the default value. You can experiment with different values to achieve the desired stylization effect.
  5. Zoom Out functionality: Midjourney 5.2 introduces a new zoom-out feature that lets you zoom out generated images. You can select different zoom levels, such as 1.5x, 2x, custom zoom, or make the image square. This feature is particularly useful for exploring complex scenes and capturing more detail by zooming out.
  6. “Shorten” command: The new “Shorten” command in Midjourney 5.2 helps you analyze the hint and suggests words that may not contribute significantly to the result. Using the “Shorten” command, you can identify unnecessary words or tokens in your prompt and exclude them to refine the result.
  7. “High Variation” mode: The “High Variation” mode, which offers more diversified results, is limited to Midjourney 5.2 and is not available in version 5.1. If you prefer to use version 5.1, you can switch to it in the settings.

Previous model: Midjourney V5.1

The Midjourney V5.1 model, released on May 03, 2023, is a significant improvement over the V5.0 model. To use it, add the parameter –v 5.1 at the end of your order or select 5.1️⃣ MJ Version 5.1 in the settings. This template offers improved image quality, more creativity and flexibility in your prompts. For example, a short prompt such as “dog” generates highly creative images and non-standard prompts such as “Aurora Champion” result in stunning artwork. However, problems remain with weapon-related prompts, for example, “futuristic bow”.

Previous model: Midjourney V5

The Midjourney V5 model, released on March 15, 2023, was the most advanced model when it was released. To use it, add the parameter –v 5 at the end of your order or select 5️⃣ MJ Version 5 in the settings. This model has very high consistency, excels at interpreting natural language commands, offers higher resolution and supports advanced features such as repeating patterns with –tile. This model is also much better at creating hands and smiles, where previous models often offered hands with 4 or 7 fingers for example.

Previous model: Midjourney V4

The Midjourney V4 model is based on a brand new code base and a new AI architecture, designed by Midjourney and trained on the new Midjourney AI supercomputer. This model has a very good knowledge of creatures, places, objects, etc. It is particularly good at details and can handle complex commands with multiple characters or objects. Version 4 supports advanced features such as image commands and multiple commands.

Version 4 styles 4a, 4b and 4c

The Midjourney Version 4 template offers three slightly different “flavors” with slight variations in stylistic setting. Experiment with these versions by adding –style 4a, –style 4b or –style 4c to the end of a V4 command.

Style 4c is the default style and does not need to be added at the end of the command.

Previous templates

You can access previous Midjourney templates by using the –version or –v parameter, or by choosing a template version with the /settings command. Each template has its own strengths for handling different types of images.

Niji model

The Niji template is a collaboration between Midjourney and Spellbrush, specially designed to produce anime and illustrative styles. This template is especially good for action scenes and character-driven compositions in general.

Latest Niji Template: Niji Version 5

The Niji Version 5 template is the latest and most advanced Niji template. To use it, add the –niji 5 parameter at the end of your order or select “Niji version 5” in the settings. This model is sensitive to the –stylize parameter. Experiment with different stylize ranges to refine your images.

Test templates

From time to time, new templates are released temporarily for the community to test and give feedback on. There are currently two test templates available: –test and –testp, which can be combined with the –creative parameter for more varied compositions.

How to switch between templates

To switch between the different templates, you can use the –v, –test, –testp, –niji parameters and add them at the end of your command. You can also use the /settings command and select the desired version in the menu.

Midjourney offers a wide variety of templates and styles to meet all needs for AI-generated images. Feel free to experiment with the different versions and settings to create unique and stunning images.

How to make a good “Prompt” with Midjourney?

In this section, we will explore the different versions of Prompts you can use to create unique and captivating images with the Midjourney Bot.

Reminder: What is a Prompt?

A Prompt is a short phrase or word that the Midjourney Bot interprets to produce an image. The Bot breaks down words and sentences into smaller elements called “tokens”, which are compared to its training data to generate an image. A well-designed Prompt will help you create images as close to what you want to create.

Important: Prompts must be written in English to be understood by Midjourney. If you don't know English well, you can use a translation service like Google or Deepl.

Structure of a Prompt

  1. Basic prompts: A basic prompt can be as simple as a single word, a phrase or an emoji. Example: /imagine: Japanese house with a temple in background and a bird on the roof.
  1. Image Prompts: Image URLs can be added to a Prompt to influence the style and content of the final result. Image URLs are always placed at the beginning of a Prompt. I will take an image created with the previous prompt and add details. Example: /imagine: https://s.mj.run/H1GaIFaxPYs two birds on the roof
  1. Advanced Prompts : More advanced Prompts can include one or more image URLs, multiple phrases and one or more parameters. Example: /imagine: https://s.mj.run/H1GaIFaxPYs https://cdn.discordapp.com/attachments/1078680728008667300/1102578285583536138/176274_Kyoto.jpg sunset light

The Parameters in Midjourney

The parameters modify the way an image is generated. They can change the proportions, the image quality and many other things. The parameters are placed at the end of the Prompt. More details in the next section.

Notes on Prompts

  • Prompt length: Prompts can be very simple. A single word (or even an emoji!) is enough to produce an image. However, more descriptive prompts will give your image a unique look. Too long and they may not be better. Focus on the main concepts you want to create.
  • Grammar: The Midjourney Bot does not understand grammar, sentence structure or words like humans do. Word choice is important. Prefer more specific synonyms. Instead of “big”, try “gigantic”, “huge” or “immense”. Reduce the number of words to give each word more weight. Use commas, parentheses, and dashes to organize your ideas, but be aware that the Midjourney Bot will not always interpret them reliably.
  • Focus on what you want: It's better to describe what you want than what you don't want. If you ask for a party without a cake, there will probably be a cake in the picture. If you want to make sure that an object does not appear in the final image, try using Advanced Prompts with the –no parameter.
  • Think about the important details: Anything not mentioned may surprise you. Be as specific or vague as you like, but keep in mind that anything you don't specify will be random. Being vague is a great way to get variety, but you may not get the specific details you want.

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

  • Subject: person, animal, character, place, object, etc.
  • Medium: photo, painting, illustration, sculpture, doodle, tapestry, etc.
  • Environment: indoor, outdoor, on the moon, in Narnia, underwater, the Emerald City, etc.
  • Lighting: soft, ambient, overcast, neon, studio lights, etc.
  • Color: bright, muted, brilliant, monochromatic, colored, black and white, pastel, etc.
  • Atmosphere: serene, calm, frenzied, energetic, etc.
  • Composition: portrait, close-up, bird's eye view, etc.
  • Use collective terms: Plural words leave a lot to chance. Try to use specific numbers. “Three cats” is more precise than “cats.” Collective terms also work, such as “flock of birds” instead of “birds.”

You can also choose to create an “In the style of” image, by adding the name of an artist, a famous photographer:

Example photographer: /imagine: man from brasil, photo by David LaChapelle –v 5

Example Artist: /imagine: man from brasil, painted by picasso –v 5

By choosing the right words, phrases and parameters, you can create real works of art. So experiment and have fun with the different versions of Prompts to create fascinating and unforgettable images!

Midjourney settings: how to optimize your image creations

Midjourney settings are options added to prompts to modify the way an image is generated. These settings can change the image format, switch between different versions of Midjourney templates, choose a different Upscaler, and much more. The parameters are always added at the end of a prompt, and it is possible to add several for each prompt.

Example of the use of Midjourney parameters:

Prompt: Housse with a purple pool –chaos 50 –ar 16:9 –quality 2

We see that the “Chaos” parameter has generated 4 quite different images (orientation, style, …)

Basic parameters

Aspect ratio: Use --aspect or --ar to change the aspect ratio of an image generation.

Chaos: With --chaos <nombre 0-100>you can change the variety of results. Higher values produce more unusual and unexpected generations.

No: The parameter --no is used for negative suggestion, for example --no plantes will try to remove the plants from the image.

Quality: Use --quality <.25, .5, 1, ou 2> or --q <.25, .5, 1, ou 2> to determine the rendering time and quality of the image. The default value is 1. Higher values cost more and lower values cost less.

Repeat: Use --repeat <1-40> or --r <1-40> to create multiple jobs from a single prompt. Repeat is useful for quickly restarting a job multiple times.

Seed: With --seed <entier entre 0-4294967295>you can specify a starting number for generating the initial image grids. Using the same seed and the same prompt will produce similar images.

Stop: Use --stop <entier entre 10-100> to end a job in process. Stopping a job early may result in less detailed and blurred results.

Style: Use --style <4a, 4b, ou 4c> to switch between versions of the Midjourney Version 4 template, and --style <expressif, ou mignon> to switch between versions of the Niji Version 5 template.

Stylization: The parameter --stylize <nombre> or --s <nombre> parameter influences the intensity of the default Midjourney aesthetic style applied to jobs.

Uplight: Use --uplight to select a “light” alternative upscaler when selecting the U buttons. The results are closer to the original grid image, with less detail and a softer image.

Upbeta: With --upbetayou can choose an alternative beta upscaler when selecting the U buttons. The upscaled image has significantly less detail added.

Template version settings

Midjourney regularly releases new template versions to improve efficiency, consistency and quality. Different templates are better for different types of images.

Niji: Use --niji for an alternative template focused on anime style images.

Example with already generated images and the “niji” parameter: /imagine https://s.mj.run/H1GaIFaxPYs https://s.mj.run/RPcc7eHn62s sunset light –niji

High definition: With --hdyou can use an early alternative model which produces larger and less consistent images. This algorithm can be adapted for abstract images and landscapes.

Test: Use --test to use the special Midjourney test pattern.

Testp: Use --testp for the special Midjourney test pattern focused on photography.

Version: Use --version <1, 2, 3, 4, ou 5> or --v <1, 2, 3, 4, ou 5> to choose a different version of the Midjourney algorithm. The current version (V5) is the default.

Upscaler parameters

Midjourney first generates 4 low resolution images for each Prompt. You can “upscale” the image of your choice to increase the size and add additional details to the image of your choice. Several upscaler templates are available to enhance an image.

Uplight: Use --uplight for an alternative “light” upscaler. The results are closer to the original grid image, with less detail and a softer image.

Upbeta: With --upbetayou can choose an alternative beta upscaler. The upscaled image has significantly less detail added.

Upanime: Use --upanime for an alternative upscaler specially designed to work with the Midjourney –niji model.

Other settings

These settings only work with specific older Midjourney templates:

Creativity: Use --creative to make the test and testp templates more varied and creative.

Image weight: With --iwset the weight of the image relative to the weight of the text. The default value is –iw 0.25.

Sameseed: Use --sameseed to make all images in the initial grid use the same starting noise and produce very similar generated images.

Video: Use --video to record a progress video of the initial image grid being generated. React with the ✉️ emoji on the finished grid image to have the video sent to you in a private message. –video does not work when enlarging an image.

Obsolete settings

Some parameters have been replaced or are no longer used:

  • -width and –w (replaced by –aspect)
  • -height and –h (replaced by –aspect)
  • -fast (replaced by –quality)
  • -vibe (now known as V1)
  • -hq
  • -newclip
  • -nostretch
  • -old
  • -beta

The Midjourney settings allow you to customize and optimize the image generation according to your needs and preferences. Do not hesitate to experiment with these parameters to obtain the best possible results.

The Midjourney subscription packages

Midjourney offers three subscription formulas to meet your needs. Choose between a monthly or annual payment to benefit from a 20% discount. Each package includes access to the Midjourney members' gallery, the official Discord, the commercial terms of use and much more.

To subscribe:

  • Use the /subscribe command to generate a personal link to the subscription page.
  • Go to Midjourney.com/account
  • Select “Manage Subscription” from the sidebar when you are logged into the Midjourney site.

Comparison of the formulas :

  • Free trial: no monthly or annual subscription cost, 0.4 hours of fast GPU time for life.
  • Basic plan: $10 per month or $96 per year ($8/month), 3.3 hours of fast GPU time per month.
  • Standard Plan: $30/month or $288/year ($24/month), 15 hours of fast GPU time per month, unlimited relaxed GPU time.
  • Pro Package: $60/month or $576/year ($48/month), 30 hours of fast GPU time per month, unlimited relaxed GPU time, stealth mode.

The generated link is unique for each account and should never be shared with others!

Accepted payment methods:

Only payment methods supported by Stripe are currently accepted, such as credit or debit cards issued by services like Mastercard, VISA or American Express. Google Pay, Apple Pay and Cash App Pay are available in some regions. PayPal, bank transfer and similar methods are not supported.

Other information about Midjourney paid accounts:

Managing your subscription :

Manage your subscription at https://www.midjourney.com/account/.

Monthly renewal :

Unused monthly fast GPU time does not carry over to the next month. The fast mode is automatically reactivated when the subscription is renewed.

Upgrading:

You can upgrade or downgrade at any time. When upgrading, you can choose whether to upgrade immediately or at the end of the current billing cycle. If you choose to upgrade immediately, you will be quoted a pro-rated price based on the usage of the plan you are upgrading from. Downgrades are always effective at the end of the current billing cycle.

Cancellation:

Go to https://www.midjourney.com/account/ to cancel a subscription at any time. Cancellations are effective at the end of the current billing cycle. Subscription benefits, such as access to the Community Gallery, are available until the end of the current billing cycle.

Refunds:

Refunds are available for subscribers who have used less than 1% of their monthly GPU minutes, including time used in relaxed mode. If you are eligible for a refund, the dialog box will automatically appear when you click to cancel your account.

Buy more fast GPU hours:

If you run out of fast GPU time before your monthly subscription renews, you can purchase additional fast GPU hours at a rate of $4/hour. Hourly rates are experimental and subject to change.

Purchase more fast GPU hours on your Midjourney account page.

Purchased GPU fast hours do not expire, but you must have an active subscription to use them.

Information about your account:

Use the /info command to view information about your account, such as the remaining fast GPU time for the subscription period, lifetime usage statistics, the number of jobs queued or running, and the subscription renewal date.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *