I mostly do oil painting, coherent (including on those two tests).
I had no idea. Wow.
Why would they leave such a big loophole? (Screencap is your friend.)
You don’t even need screen cap. Just right-click, save image as. Or even “copy link to image” but who knows how long that lasts. As long as you don’t log in, it doesn’t seem to matter. I first just went to different devices in the house for extra play time (like my iPad, then my kids’ iPads) and then I realized, wait, that’s just cookies then, isn’t it? Seems so.
Yeah, just tried that. Did “He saw a picture of a lion in a blue cloak, standing off a cliff in the sky, at his side.” (from my AI generated text) and even upscaled it to 8000x4800.
Art Quiz!
Assign the following artists to the depiction of the Easter Bunny
Banksy
Boris Vallejo
Caravaggio
Frida Kahlo
Georgia O’Keeffe
Norman Rockwell
Pablo Picasso
Pierre-Auguste Renoir
Answers
Kahlo, Caravaggio, O’Keeffe, Banksy
Renoir, Picasso, Rockwell, Vallejo
Catherine of Aragorn
Keywords: [Katherine of Aragorn oil painting by James Gurney renaissance painting artwork beautiful oil on canvas polished]
Those eyes… those eyes…
This thread needs a “like” button.
“The Walk” from The Wild Bunch. Original screen shot:
And “the creation”:
It made me giddy that the software understood “You’re gonna need a bigger boat.”.
Thanks for the tip about incognito.
I used it to create two images with the exact same prompts and seed value.
They did NOT produce the same image, although they were pretty similar, so there is other randomness going on besides the seed value.
Wow, the AI did that? That’s a great job on the face. The hand is a bit off, and I’m not sure why she’s holding a whisk, but it’s still quite impressive.
What I don’t understand is that the AI doesn’t seem to know that the eyes of a human are usually almost symmetrical, but these look like eyes from two different people. Shouldn’t that be an easy rule to teach to the AI or to learn it itself, respectively?
It kinda knows what a TARDIS is. I did specify Police Box to steer it more in that direction, though.
So here’s a thing I did.
1.) I literally laughed out loud when I saw that the AI actually drew Godzilla visiting a grave
2.) It looks like having the word “strange” in a prompt is a good thing.
Also, absolutely nobody could ever guess, given a million years, what prompt generated this.
Perfectly cromulent.
I mean, W the actual F?
Obviously that one just has a different setting on its chameleon circuit.
This is the original. The AI made her bustier and necklace look like a buckler without a protective hub.
A few years back there was a meme/fad of people posting
Danbo | Know Your Meme based on a character (actually a costume) from a chapter of a manga. I gave the name (Danbo) a try and added the manga name (Yotsuba) to give a little help. The resulting image gave definite impressions of trying to draw small, boxy robots (if you know what the prompt was) but what really impresses me is that it generated a pretty darn clear version of a character from the manga.
It knows Grogu by name. It knows The Mandalorian by name. It seems to know all sorts of fairly obscure and/or recent pop culture items. Not the type of things I would think would be selected for a hand-tuned training set.
“Bon Scott downing a pangalactic gargle blaster”, psychedelic. Weird (I didn’t know that the blaster came in bottles). Is this Zombie Angus on the right?
Holy crap holy crap holy crap holy crap.
Has nobody else tried doing video yet?
I have always used the “coherent” algorithm instead of the “artistic” one and it doesn’t give the video option, but I happened to look at the options with “artistic” on.
This is my tried-and-true “yokai parade” prompt.
5 seconds and 750 iterations is the best you can get with 5 credits, but you could do a lot more with a lot of credits to throw at it.
(Reversed and looped by me in post.)