Some time ago I wrote a story called Julia and the Red Fox, featuring a red fox. The fox appeared to Julia inside a dream. I enjoyed writing this tale because of the lucid imagery that I had to conjure up. I even sketched a red fox outside the window with some markers and used it as a cover.
“Well?” said the red fox peeking through the window. “Are you coming or what?”
This masterpiece of human creativity has not been rivaled by any artist since then, but I'm afraid the time has come to pass on the crown to a new generation.
The other evening I thought it'd be jolly good fun, as the Brits say, to generate a red fox based on the story. So I quickly fired up Stable Diffusion (SD) and began prompting the AI to generate an image as follows:
a red fox outside the window like an intricate dream with meadows and beautiful sunset, there is a river in the distance, delicate smooth fields with swaying grasses, the sky is psychedelic.
There is more to the prompt, but I won't embarrass myself in public more than I already have. So, I will keep it a secret.
Could it be possible, I asked myself, that a machine could rival the creativity of man? I know that my red fox will not hang at the Louvre because DA MAN is in control of things, but you have to admit, that the illustration has a certain je ne sais quoi. I think it's the psychedelic sky. Those swirls always make stuff look alright. Like my sweet dear grandma used to say, when in doubt, go psychedelic.
So, the question remains, could it be possible for a machine to outshine this work of genius? Let us see the five images generated and discuss!
Not bad. There is a red fox, a field of flowers, but no window, river, or psychedelic sky. Sorry SD, this round goes to chez moi.
Oh look at that cute little fox. Nice smooth grasses and kinda psychedelic sky. Great positioning by the AI (somebody has been watching youtube videos on composition). Sadly, cuteness doesn't get you any points. There is no window and no river. So round two goes to ME!
Is that a fox or a cat or a FoxCat? The sky barely has a hint of psychedelic, the trees and grasses do have a nice dreamy quality. There is a window frame! Very close. I just can't get over the look of that fox though. So this is a tie between SD and me.
Ooo. Check it out. Now we're talking. A bit of a surreal sky with soft grasses (that I assume are swaying), a window frame, and Julia's bed! No river but wow, that's very close to what I had in mind. So this rounds definitely goes to SD. Booo!
"Oh hey! Are you asleep or are you awake?" This is a wonderful creation. It is missing the river, which didn't appear in any of them, but the fox looks alert (not creepy like the previous one), the grass is smooth with hints of swaying, and you can see a faint sun in a pinkish sky. Very nice. I'm man enough to concede that this round goes to SD.
Let's not kid ourselves here, even though the generated images are relatively simple, the AI kicked my butt in two minutes, which is basically the time it took my computer to create the five pieces from the moment I hit enter. Without a doubt a real generative artist could create even more spectacular and detailed creations with some elegant prompt engineering. Of course, the AI needs to be prompted and cajoled into making an image by a human operator. Won't be long before it doesn't need our input and just learns to create art for the sake of art. Some artists are doing a lot of soul searching with the introduction of this new medium, and some of us are already working in the sugar caves of our new AI artist overlords.
Images generated by 