It takes a second to get used to, but once you do, MidJourney is a powerful tool. It sounds more complicated than it is. You start at their website at MidJourney.com:
Click “Join the Beta” and you will be taken to their Discord server. This is where you’ll do all your art generation by interacting with the bot.
Once you join, go back to the website and click “Sign In with Discord.”
This is where you will get access to your gallery. When starting out, this will become invaluable to go back and reference your work, as those public channels move fast.
Now you’re setup and ready to go. If you’re unfamiliar with Discord, each server has channels listed on the left (you may need to tap the three-line icon on the top left of the screen to see them on your phone)
Scroll down a bit and you’ll see the bot channels they set up. The bot channels are the ones listed under Newcomers (where they recommend you start), General Image Gen, and Themed Image Gen. Feel free to browse any available channels to watch what everyone else is doing and learn.
When you’re ready to go, join a channel and start creating.
Or you can choose to DM the bot or set up your own Discord AI art server to invite MidJourney’s bot, along with friends and other Generative AI bots.
Like any Discord bot, you access available commands by typing /. This will show you the extensive list of commands available for both MidJourney’s bot and Discord. To generate an image, type /imagine and press space. This will open the command prompt for you to tell it what you want to draw.
From here, you have a wide array of options and can be as specific or vague as you like. There is no right or wrong answer, and you can get wildly different effects by changing things up.
Image prompting is where the magic happens, and it’s kinda like learning to play an instrument. You can give it something like “an apple” and just roll the dice, or you can specify “an apple in the style of Van Gogh” or “detailed pencil drawing of an apple” or even “a person eating a metallic purple apple” to get what you want.
However, it’s not so simple, because AI language syntax differs from the way we typically speak. Here is the high-level syntax directly from MidJourney’s documentation online.
All of this is optional, but if you use the elements, they should follow this order. At the start, you can link directly to any images you want the AI to specifically consider when creating. Make sure you are linking directly to the image file, or you will get an error.
Next, you type whatever words you want, and you can weigh them by changing the order, using capitalization, and using the command ::w.
Finally at the end, input any specific parameters you want like image height, width, and aspect ratio, uplighting, whether you want it to record the generation as a video, etc.
Click that link above to the prompting guide because it is essential and covers so much more than I will here.
Once you input what you want and send the command, it will begin the job and you’ll get 4 output images in a 512×512 pixel square.
You will see buttons below labeled U1-U4 and V1-V4. I labeled the numbers above, and U=upscale specified image to 1000 pixel square and V=create variations of the specified image. Or you can press the 🔄 button to reroll and generate 4 completely new images based on the same prompt.
And of course, you can always start over from scratch.
If you like the image enough, you can max upscale it to a 1600 pixel square. These are the basics. Once you get into it, you can figure out seeding and using the bot via DM or your own server, light upscaling, etc.
For what seems like a simple thing, it’s pretty complicated and intricate. But like a guitar, there are varying skills levels and styles that can achieve something magical.
There are also models you can choose. You access them (and other settings) by typing /settings.
MidJourney Discord settings
Each mode will provide a completely distinct result. V4 is the most current as of this writing, but you can also choose from versions 1-3, test, testp, and niji modes. Try each one to see the vast differences in results.
If you want to see you usage, type /info
If you love it enough (or have monetized it like I have), you can type /subscribe to get a unique link to buy a monthly or annual subscription.
That’s pretty much all there is to it. You can be vague or super specific and get some insanely cool things out of it. Here are a few examples from my first month in v3 to get your creative juices flowing.
Cthulhu as a Beanie Baby
A Spider-like orchid
A snowflake on fire
Mike Tyson infected with the Venom symbiote
A Pokeball aquarium
Betty White as Harley Quinn
Also published here.