How to Use Midjourney
Midjourney, an AI image generator, provides a means to create impressive imagery from plain text prompts using only our browser. It is a generative artificial intelligence program that provides its services through Discord bot commands. It entered open beta on July 12, 2022 with version 3. It has been releasing new versions periodically since, and in June, 2023 it released version 5.2.
Midjourney has seen widespread popular use, being used to write illustrations for a children’s book and generate the cover of The Economist magazine in June 2022. In one of the most notorious examples, it was used by Jason Allen to create an image he named Théâtre d’Opéra Spatial which took first prize in the digital art competition of the 2022 Colorado State Fair.
Let’s take a look at what we can do with Midjourney. The following images were generated with fairly simple prompts.
This picture used the prompt: “Happy photorealistic boxer dog fetching a ball”:
For the following picture, the prompt was: “Old antique console radio”:
For this one, the prompt was: “Rusty old warning sign”:
In this tutorial, we’ll learn how to use Midjourney, and how we can get started creating and manipulating our own images.
Subscribing to a Midjourney Plan
The first thing we need to do to work with Midjourney is to sign up for a plan. There are four tiers that offer different features. The primary difference is the amount of GPU time we get per month to produce images. The two highest tiers provide us the ability to use a “stealth” mode meaning the generated images aren’t available to all Midjourney users (the default). Each tier is billed monthly, but there is a discount if we purchase a full year in advance. As of this writing, the tiers and pricing are as follows:
- Basic Plan: $10 USD/Month. 3.3 hour/month GPU time.
- Standard Plan: $30 USD/Month, 15 hour/month GPU time.
- Pro Plan: $60 USD/Month, 30 hour/month GPU time.
- Mega Plan: $120 USD/Month, 60 hour/month GPU time.
Details on each plan can be found on Midjourney’s website.
To sign up for a plan, you will need a verified Discord account. If you don’t have one, you can find instructions on Discord:
Once you have a verified Discord account, visit Midjourney.com:
- Click the “Sign In” link in the lower right of the Midjourney landing page.
- You will get a page asking for the Midjourney bot to have access to your Discord account.
- Click “Authorize” and you will be brought to a page where you can purchase a subscription.
- Make sure you select your preference for monthly or yearly billing and click the subscribe button for the plan you want.
- You’ll be taken to a page where you can fill out your financial information and subscribe.
To use the Midjourney bot after you’ve signed up for a plan, you will need to join the Midjourney server on Discord found here: http://discord.gg/midjourney.
Note: You can find out about joining Discord servers on the Discord website.
Now that you have a plan and have joined the Midjourney server, you can start creating images.
Your First Prompt
To create new images with Midjourney, you send a message to the Midjourney bot using the following process:
- Log into Discord and navigate to the Midjourney server.
- Go to any channel named “newbie-#” or “general-#”, these are the channels used for image generation.
To generate a set of images, we begin with a Discord bot command. Discord bot commands are prefaced with a “/“. The command we use for image generation is “/imagine”. After “/imagine” the word “prompt” will appear, which is waiting for us to enter the text description of what image we want to generate.
When we send this message, the bot will respond with a set of four images based on the prompt. It might take a few seconds, but the bot will show its progress as it is generating the images.
Note: Midjourney works best with short descriptions of what we want to see. Instead of “Show me a picture of a Renaissance oil painting showing a knight wearing armor on a horse.”, use something like: “A Renaissance oil painting of an armored knight on horseback.”
The simple prompt above created the following image:
In our prompts, we can specify an artistic medium or style for our image. Here are a few examples:
Prompt: “Watercolor of an armored knight on horseback”:
Prompt: “A Medieval illuminated manuscript page of an armored knight on horseback”:
Prompt: “A coloring book page of an armored knight on horseback”:
Prompt: “Graffiti art of an armored knight on horseback”:
Another thing we can add to our prompts is emotion. Adding an emotional cue to a character can give a more expressive image, as in the following:
Prompt: “Happy dog”:
Prompt: “Sad dog”:
Prompt: “Surprised dog”:
Prompt: “Tired dog”:
We can also influence the image by specifying a color scheme:
Prompt: “Crimson cow”:
Prompt: “Blue cow”:
Prompt: “Black cow”:
Prompt: “DayGlo cow”:
Saving Your Work
To save your images:
- Click on the image.
- Click the “Open in Browser” link.
- In the tab that opens, right-click on the image and select “Save Image As…”.
Note: Opening the image in the browser gives us the full-size images, the images we see in Discord are scaled down to appear easily on the screen.
Midjourney keeps a personalized gallery of all our images. This is useful in case we lose track of our creation in a Discord thread. To recover our images, we go to https://www.midjourney.com/app/ and sign in with our Discord account.
Reroll, Variations, and Upscaling
There’s more we can do with an image after we initially create it. You may have noticed that in addition to the four images created by our prompt, there are several buttons under the thumbnails:
These allow us to perform various tasks with the images we create.
The first thing we can do is reroll with the button on the far right:
Rerolling generates four more images from the same prompt. This is helpful when we think the prompt is good, but we don’t like the first four images or want some more choices. Here’s a reroll of the images of “a young girl walking a dog” above:
We can also create variants of an image that generates four new images using the selected image as a base. We do this by selecting “V1”, “V2”, “V3”, or “V4”. Each one will create variants of one of the four images. The images are numbered from the top with the upper left being 1, upper right being 2, lower left being 3, and lower right being 4.
Here is a set of variations based on image 2 of our reroll above (Using the “V2” button):
We can also upscale an image. Upscaling an image takes one of the four images and displays it by itself for further modifications. It’s called upscaling because prior versions of Midjourney produced four low-resolution images and it was necessary to upscale to get a full-size image. In current versions of Midjourney (5 and above) the original four images are already full-size, and upscaling just crops out one of the images to work on.
We upscale an image by clicking on the “U1”, “U2”, “U3”, or “U4” buttons. Each one will upscale one of the four images.
Once we upscale an image, there are even more things we can do with it.
Manipulating an Upscaled Image
Midjourney provides more options once we upscale an image. Here is what we see upscaling image one of our “girl walking a dog” reroll:
There are more buttons under the image now. The first two, “Vary(strong)” and “Vary(Subtle)”, are just like the “V1” through “V4” buttons earlier. They both make four new variants of the image. The first button, “Vary (strong)”, makes significantly bigger variations than the second button, “Vary (subtle)”.
The “Vary(Region)” allows us to create variants that change only a part of the image. When we click on this button, an editor window pops up:
There are two buttons in the lower left that change how we select an area of the image.
The rectangle tool makes rectangular selections, the lasso tool allows freehand selections. We click on the tool we want and click and drag over the area in the picture we want to make variations of.
Once we’ve selected the area to vary, we click the “Submit” button on the lower right. This will produce four more variants with changes only in the selected area of the image. Doing this with the selection above gives the following variations:
Under the “Vary” buttons are the “Zoom Out” buttons:
These will shrink the image and fill out the margins to varying degrees. Here is our upscaled image after “Zoom Out 2x”:
The “Custom Zoom” button opens a popup box where we can edit the prompt with a custom zoom amount. Just modify the number after “–zoom” to a value between 1.0 and 2.0.
Below the zoom buttons are the panning buttons:
Each one will pan the image, expanding it in the direction the arrow is pointing. Here’s the result of using the pan-down button on our upscaled image:
This is useful for shifting the focus off the center of the image, as Midjourney will usually initially place the focus of the image in the center. By strategic zooms and pans, we can alter the composition of the image.
There are two more buttons we see after upscaling:
The heart allows us to mark the image as a favorite, easily referred to on our Midjourney profile page by selecting the list of favorited images. The “Web” button will take us to the upscaled image on our profile page.
This tutorial covered the basics of using Midjourney. We went over signing up for a Midjourney account and using the Midjourney bot. We learned the basics of crafting prompts to generate images. We also saw how to manipulate the images we generate by using rerolls, variants, and upscaling, as well as manipulating an upscaled image by creating variants, zooming, and panning. With this knowledge, you’re equipped to start making your own images with Midjourney.
To explore additional topics on generative AI, check out some of the articles located here: AI Articles.