Image-generating AI roundup: DALL-E, Craiyon & Midjourney
Look, I know it says “The Daily Typographic” on the tin, but if you carefully read the ingredients list (to keep the metaphor going, I mean the About page) it says
Crawling the web for new, exciting typefaces, sites, brands and type-adjacent content.
And since, let’s face it, anything can be type-adjacent, including Italian cheese (I mean, goddamn it, parmesan even has stuff printed on its rind) and toilet seats (um, even I know about about Kohler) I can pretty much write about anything. CALL ME THE MASTER OF LOOPHOLES.
If you’re following me on Instagram (and if you aren’t, why u make me sad?), you may have noticed that I’ve been using more illustrations as backgrounds for my type designs. Those are not stock illustrations and are clearly not made by someone who cannot draw a stick figure without fracturing his collarbone (me), so where do I get them?
Generative AI software, of course, and more specifically, Midjourney.
The concept of art-generating AIs
Generative art AIs are artificial intelligence models trained with the express purpose of being able to generate images from text prompts. There are a few of these around, but only two are readily available.
Google’s Imagen AI is not available to plebs like you and me (but mostly you) so there’s not much to say about it.
OpenAI’s DALL-E 2, meanwhile, is currenly in beta, but they still haven’t let me in, citing the stench of my armpits as the main reason. They have suggested that I take a bath, but at this point I don’t want to remove the thick protective layer of grime that has built up on my skin in case World War 3 happens soon.
These leaves us with two readily available AIs.
Craiyon (formerly DALL E mini)
Craiyon was the first generative AI I used, and I was immediately in love. It is not perfect — far from it — but as a showcase of what can be done with this sort of software, it was pretty wild to see the generated images.
The images are currently very low-res, and their quality simply does not compare with the results you can get from something like DALL-E 2 and Midjourney, but as a concept… oh, man!
At its current state, Craiyon is not something that I’d use in a design, as the results, in addition to being low-res, can be pretty garish. However, as a first glimpse into the world of generative AI art, it can be mindblowing. And if all you want is to generate some funny sight gags, such as “Shark Hitler shooting laser from his eyes” or “Shub-Niggurath having a drink at a bar”, it’s just *chef’s kiss* mwah!
Midjourney
Midjourney is… quite honestly, Midjourney is life. Well, at least my life for the past few days.
What elevates this particular AI well above the level of Craiyon is the quality of the results and the fact that it renders at a much higher resolution; high enough for the results to actually be usable in creative projects.
Here’s something I made today, followed by the prompt I used to create it
The prompt for this particular example was “futuristic city with pointy buildings with retro esthetic :: black :: grey :: white :: extreme halftone :: risograph :: xerox”
As you can see, the prompt can contain directions for scene creation as well as dictate the stylistic choices you want the AI to apply.
Midjourney generates four distinct images to start with, and you can ask it to create variations on those images until you’ve reached the desired result. I went through four iterations to reach what you see above.
In contrast to Craiyon, which is currently free and allows unlimited image creation, Midjourney will (at the time of writing) let you create up to 25 images with a free plan. Then you can pay $10/month for 200 minutes of server time, and there’s also a $30/month plan for unlimited image creation. Best $30 I ever spent.
Here are a couple more Midjourney images I made earlier this week
Conclusion
You may be thinking that talking about image generation AIs was a strange detour for the blog to take, but I did it for two reasons:
I’m super excited about the technology, and I firmly believe that it’s here to stay. Normies won’t get on the train for a year or so is my prediction, but after that it’ll blow up. That’s also why me and a friend are releasing a book very soon about the subject, which you can preorder here. Nothing like a soft sell.
This is a design site, and I’m pretty sure that it’s actually designers that can take full advantage of something like this. My aunt Mauve doesn’t know what risography is, but you do!
I’ll probably be returning to the subject of art creation AIs in the near future. Until then, have some ice cream and get a tan. Oh, and get my brand-new book “Create Stunning AI Art Using Craiyon, DALL-E and Midjourney”, out now on Amazon!
Download 20 gorgeous pages from Clothes for Language, my irreverent typography guide. Click here to get it immediately. No signup necessary.
