A few months ago I saw a link to a website called NightCafe that lets you type in a text prompt and an AI-model (VQGAN+CLIP, if that means anything to you) will turn it into art. A lot of the results are actually kinda awesome, especially if you don't have a specific image in your mind and are just playing around with interesting prompts. The way the AI creates the image has this sort of surreal, organic fractal quality to a lot of the work, while also having some sort of greater "artistic" quality to the output. I refer to the general style of art that this thing creates as "hyper nightmare" and I think you will agree with that summation when you see some of the below results.
The website itself uses a credits system for making images; create an image at the cost of 1 credit. Create an image with better resolution, 2 credits. Create an image with increased algorithm iterations, 4 credits. You get the drift. You obtain some free credits on sign up and get some extras for doing some basic actions on the site, such as liking a certain number of other creator's art, publishing your own creations, sharing work on Twitter/Instagram, etc.
I played around with a few ideas, thought it was somewhat interesting, and then completely forgot about it and didn't visit the site for three months.
While working on another project that involves content-creation via AI, I remembered NightCafe and returned to it, hoping to coax the benevolent algorithm to create interesting looking images from my lackluster descriptions. I ended up buying credits because I didn't want to wait as I fine-tuned what I needed for the other project, and because I've spent a lot more on a lot worse before. I'm pretty happy with the service overall, and I had a few leftover credits so I decided to take a stab at using some of my Neocitizen friends as the prompt keywords to create some robotic-crafted art. Without further ado, here we go:
Not the best result, to be honest. This is the type of "dud" result you get when your word prompt is something ambiguous, or overly abstract. It does give the impression of lots of books though, set in a sort of never-ending nightmarish avalanche, which is exactly how it feels to visit TRP and realize they've reviewed over 450 books and are still going. What the heck?
I hear tell that The Reading Project has created a Youtube channel as well; their colony expanding ever-outward, creeping into all spaces of the internet until they take over everything and all that is left, the only thing that exists, one thing and all things: The Reading Project.
I also like how it vaguely, vaguely gives the impression of someone sitting there, reading a book. Or, that's how I view the image at least. Art, am I right? No, for real; am I right? Is this how I'm supposed to interpret the image? Someone validate me. Please.
Oh, drat. Another underwhelming creation. This little robo-painting-machine really does struggle with abstract names or ideas. But still, you can see where some of the intention lies, if algorithms have intentions, anyway. We see some sort of language/learning puzzle on a faux-wood desk that is well lit by overhead fluorescent lighting, reminiscent of those school days gone by.
I'm not one to necessarily equate "school" with "learn" immediately anymore, but obviously this is a first connection most will make. I just wonder how this image creator decides to create what it creates based on a few words as a prompt. Maybe we'll never know. Maybe we aren't supposed to know.
Whoa, now this is some fucking Art. Capital "A" and everything. Look at this shit, we got two angels, descending. Perfection. It also appears that they're descending books that are as big as and/or are literally mountains.
Does this represent the descent of man from a place of pure knowledge, downward into the depths of mortality as we accept that we will never know anything in the grand scheme of things, but we must accept this inevitable fate regardless, because there is no escaping it? Robo-painter, can you stare directly into my soul and crush me with a single digital image?
Also for some reason, in the background there's a line graph going up, down, up; like that fucking stonks meme or whatever. lol, what the fuck is going on here?
Oh yeah, you can use multiple separate prompts and give them different weights for how the AI makes this magical art-thing. I gave these two prompts equal weight, as if they were strangely entangled together, almost as if they were one person, but they're not, they're two. They are definitely two different people. Two.
This image is pretty "psychedelic" or whatever, trippy and weird and wavy and good. I love the little weird ear lobe/chicken beak type of morphation near the center of the image, like what is that? Why did Mr. Robot Artist throw that in there?
P.S. This image is what your brain looks like in a 90's movie.
And here we see the automatic llama, apparently hanging out with misshapen other automatic llamas. Munching around the green and purple mountainscape, the automatic llama disobeys conventional imagery; is that side the head, or the tail? What about the other side? Why do the other llamas look the way they do?
When I used the prompt, I admittedly was hoping for the AI to create a result more like "Mecha-Godzilla," but we got this strange, distanced little llama-scape with hints of automaton llamas and impossible landscapery. I'm okay with that. The key to having fun with NightCafe, and with life in general I think, is to have very tempered expectations and appreciate the weirdness that does come your way, as unpredictable as it may be.
I have a few credits for NightCafe still remaining, so if anyone has a prompt they want to see created into AI-generated art, use the comments below to suggest some weirdness.