You will want to begin by creating a Discord account. While it’s an optional step for Stable Diffusion, it is required to interface with and to talk to MidJourney. Unless you are dead-set on only using Stable Diffusion, I strongly recommend not skipping this step.

If you ever used Slack, Discord is quite similar. The main difference is inclusion of voice channels – a feature you don’t ever need to use if you don’t want to. Just like Slack, Discord can be run using a standalone app or via web interface. Both of the options are available on any platform of choice – be it desktop, tablet or mobile. If you don’t want to install the app – web version has all the functionality you’ll ever need to successfully produce AI art.

• Simply go to https://discord.com/ and hit the Login button on the top of the page.
• If this is your first time using Discord, press the Register link at the bottom of the window and set up your account.
• Once done, go back to the main Discord page and log into it.

Once logged in, you can leave that browser tab running, or safely close it. It’s time to move onto our first AI.

• Visit MidJourney’s webpage at https://www.midjourney.com/home/ and admire a cool ASCII vortex background they created.
• When ready to proceed, hit the Join the beta button.
• A new window will open, inviting you to join MidJourney’s Discord server. Accept Invite and you should be taken to your Discord App, or webpage – depending what you chose in step one.
• Once on MidJourney Discord, use the left tab to switch between the channels and the bottom text field to type.

I strongly recommend visiting the rules, faq and getting-started channels to get a better sense of the server’s etiquette and rules. For now through we are in a hurry to get you started, so let’s keep moving!

• Enter one of the channels labeled as Newcomer Rooms. It will be one of the many channels named #newbies-XX (where XX stands for a number from 1 to 99). It does not matter which number you pick.

You will see a stream of prompts and images from other users. If the channel is too full and scrolls too fast, pick another one. Some ought to be slower and less chaotic.

You are currently in the free trial mode. You have been given a generous number of computing credits to experiment with. Once these credits run out, you will be asked to subscribe to a paid tier (it’s worth it), but for now, we’re riding free.

• Click in the bottom text field and type /imagine followed by space.

The word prompt should pop up in the text bar, asking you for what you’d like the AI to generate.

• Continue typing, describing what you’d like to see. When ready, hit Enter

I recommend starting with a relatively simple prompt and then trying it again and again, adding more detail each time. For this guide I tried: oil painting of doctor kitten by Michelangelo.

MidJourney AI will now begin dreaming up your image. It will start with a canvas filled with noise, and over several seconds continuously refine the image, adding more and more detail that matches your prompt.

If you are on a fast-scrolling channel, the generation may be swept away, but don’t worry – when finished, it will pop up at the bottom, and you will also receive a notification.

…by the way congratulations! You’ve just generated your first AI image!

A finished generation will look something like this:

Each of the four images was generated using a different random seed. Think of them as different interpretations of your prompt.

They are also quite tiny. MidJourney is able to work really fast, because at this stage it produces thumbnails only. Next then, you will want to refine and scale up your favorite(s).

The buttons below the four thumbnail images allow you to do just that:

• U1, U2, U3 and U4 are the upscale buttons. If you like one of the four images, simply press a corresponding upscale button to blow it up to large size.
  It will kickoff another generation process and take about a minute before you can download your finished artwork.
• V1, V2, V3 and V4 buttons will allow you to refine one of your thumbnails and create variations of that particular image.
• If you don’t like any of the images, you can either refine your prompt by starting with /imagine and rephrasing your idea, or hit the blue button with white arrows to try the same prompt with four more seeds.

Say I like the one at the bottom-left (number 3) but it’s not quite there. Hitting V3 will generate variations of that particular random seed. These will look very similar to the third image and keep its general composition, but tweak and randomize various details of the picture.

You may notice MidJourney auto-generated a prompt to do the variants. It includes my original text as well as a URL link to the original third picture I asked to refine. It was then able to look at the image and re-imagine it in four different ways.

I didn’t particularly like any of the new variants, but the original one was good enough so I decided to upscale that one by hitting U3 on the earlier result.

The final result will be presented to you in a format pictured above.

• You can continue to refine your image with further variations or different upscale redo options.
• If you are happy with the result, simply click on the image to see its full size – then save it to your device.
• The four emoticons below the image help further teach MidJourney AI on how well it did. As users rate their generations, the AI strives to follow methods that produce more loved and fewer hated results.

You still have a ton of free credits on MidJourney, so feel free to stick around and continue experimenting.

There is a lot to learn about prompt crafting, additional command parameters, setting up aspect ratio of your images, amount of chaos or passes AI will do while dreaming up your images.

It’s a deep rabbit hole, but you just took your first steps, so once more: congrats!

Once you feel ready to learn more, we can move on to…