I’d love to be able to see all 200 iterations instead of just 0, 50, 150, and 200. Sometimes there is a massive change in the gap between them, and it would be nice to see how it happens. (Also to make video clips of the whole process.)
You would think with the number of cat photos on the internet, the AI would have a firmer grasp of them.
I thought this was the pet forum. That’s a real photo of my cats.
Now this is a haunted forest!
Although I was originally trying for something in the style of an Annie Leibovitz photo.
And here’s another Ansel Adams:
I like this one, though I’m not sure I would call it “heaven’s gate”. Input was “heaven’s gate a.r. penck.” Oil painting, coherent, one pass:
The stick figures seem to capture his style pretty well.
My favorite results of the day:
Turning this
into this. (Looks like little carnival glass figurines.)
And this
into this. (This one preserves almost nothing of the original, and did nothing I was aiming at. But I love the result.)
(I also took an early iteration of that one and ran it as a starter image and got this, which I evolved into this.)
Bob Ross seems to be a viable style.
A different take:
I have been trying for things shaped like things, like the shape formed in frost or condensation on a window, or in blood spatter. (With mostly poor results, though “blood spatter” will give you something out of a horror movie and “condensation on glass” does give some subtile effect.) Then I tried for “marble statue of” and “woodcarving of”. I only tried those once, but they were catastrophic failures.
Then I tried “cloud shaped like” including a duck, a bat, and a screwdriver. Those were failures, too, a half-assed coating of clouds around poorly-detailed versions of the objects. Then it occured to try a normal set of mods with “style of cloud photography” added in. I went with one of my yokai portrait prompt sets:
Yokai | full-length portrait | tronie | hyperdetailed colored pencil drawing | 8k resolution | hyperrealism | volumetric lighting | photorealistic | style of cloud photography | horror Gustave Doré Greg Rutkowski
And, woah.
But I wasn’t sold on the colors, so I tried again with “colored” removed.
Woah again.
Here’s another one:
And a landscape (Japan landscape | hyperdetailed pencil drawing | 8k resolution | hyperrealism | volumetric lighting | photorealistic | style of cloud photography.)
Here are more experiments, including a few Yosemites.
Wow! You’re getting good at hacking this thing.
I make a lot of guesses about terms. Art terms, photography terms, people’s names. For instance, I mentioned loving the work of Mark Ryden, whose name has as a strong influence. Googling showed him to be considered part of the “lowbrow” art style, so I googled for other lowbrow artists, tried a few of them. Some have significant effects. For instance, “Kenny Scharf” produces broad areas of bold colors in an abstract form. Just now I’ve been loving combining him, Zdzisław Beksiński (who is listed, and whose style couldn’t be any more different) and cloud photography with my yokai diorama settings. It creates weirdly beautiful (or beautifully weird) sorta abstract images.
(I published two more on my official account tonight.)
Another interesting mod is “Tim Burton” (which is listed). It tends to make characters with shaggy black hair—when it isn’t trying to make Beetlejuice:
One of my first black and white images (using “hyperdetailed pencil drawing”) was a Tim Burton:
As is this excellent yokai portrait, which has the shaggy black hair I mentioned.
(As an aside, the yokai the AI produces are basically nothing like what I was looking for the first time I used that term, but it is still very useful.)
One thing that I’ve noticed is that the names of female artists tend to produce female-looking characters vastly more often than the names of male artists. One female artist to concider trying is Margaret Keen, who made those paintings of large-eyed kids that used to be everywhere. Her name tends to produce large-eyed (female) characters.
Semi-autobiographical:
And exactly how you remember it?
Not a bad attempt at creating virga clouds.
Got a little silly again:
Nice.
I had tried a Doctor Seuss “castle” test, but hadn’t seen an effect big enough to dig deeper. Your post made me try portraits, and apparently the AI got confused. I asked for “style of Doctor Seuss”, and it thought I asked for “style of if Pablo Picasso and Salvadore Dali got together and had a bad acid trip”.
Your second and third cloud photography images don’t look to me like a demon made of clouds, but more like a demon emerging from clouds. Still cool effects, though.
So, style of Dr. Seuss, then?