This year has seen an explosion in the quality of AI art tools. At the start of the year, I spent a long time using the open-source VQGAN+CLIP and was blown away by the variety and beauty of its output. Still, it had a lot of problems making coherent images and usually included some glitches. This summer I spent a lot of time playing with Open AI's tool Dall-E 2, which was a significant improvement on previous tools and produced some genuinely beautiful images. And now there's Midjourney v4, which can be baffling to start using if you're not a regular Discord user.
How to use Midjourney
For some reason, Midjourney decided to use the messaging platform Discord as their interface, which adds a layer of complexity to an already complex tool. To sign up, go to midjourney.com and click Join the beta. You'll then follow the steps to set up a Discord account if you don't already have one, and finally, you'll get to the Discord interface.
Once you're there, look on the left side for the newcomer rooms. and click on any of them - they're named things like newbies-53.
To generate images, click into the message box at the very bottom of the screen and start typing /imagine with the leading slash. Once you've typed this, the interface should add the word prompt and then you can type in whatever you want it to draw, for example:
/imagine prompt smiling psychedelic colorful neon cartoon monster
Hit return and you should see a new entry in the constantly scrolling messages above that says something like
If it doesn't do anything, you probably forgot to put /imagine at the beginning or didn't type the prompt in the black prompt box. Have another go.
This will almost immediately scroll off the page as there will be other users in this room, so if you want to watch your monster develop, you'll need to keep scrolling up to see it appear.
Once it's finished, usually in about a minute, a new message will appear at the bottom of the message list with your finished work. You get four images, all generated at low resolution.
The eight buttons below the images allow you to either upscale (U) which makes them bigger or generate variations (V) which creates different versions of the same concept. Let's get variations on the third one by pressing V3. Image 1 is in the top left, image 2 is in the top right, image 3 is in the bottom left and image 4 is in the bottom right.
Nice, I like the first one, let's make it bigger by pressing U1.
Looking good. I'd like to make it even bigger, so let's click Upscale to Max to make it full-size. Click the image below to see the full-size version.
Midjourney --v 4
Love it. But at the time of writing, the default is to use Midjourney's v3 algorithms, which are beautiful but not as amazing as the new version 4. According to their press release from a few days ago, it is better at handling details, has vastly more knowledge of art and can deal with more complex prompts.
To use version 4 you need to add --v 4 to the prompt, like this:
/imagine prompt smiling psychedelic colorful neon cartoon monster --v 4
That's quite a difference!
The more detail you can add to your prompts, the happier you're going to be with the results. So let's try:
smiling psychedelic colorful neon cartoon furry monster, highly detailed beautiful 3d render, made of fur, dark background --v 4 --q 2
The --q 2 option makes it higher quality, but it'll either be slower to generate or it'll cost you more if you're paying to use the system.
I'm sure you'll agree that is significantly more interesting than the v3 monster.
Comparion of v3 and v4
To give you an idea of the difference between the existing v3 system and the new v4 system, here are some example images using the same prompt. Click an image below to see it larger.
Version 3 images
Version 4 comparisons using the same prompts
Give it a go
If you can get over the clunky user interface, this tool will bring an enormous amount of joy into your life. Give it a try and share your results with us!
To see more of my art check out my Instagram @pitapatosaurus