1994 VHS screencap of death metal band on MTV Unplugged
I tried something dance related after watching the last episode of Andor.
I considered this try a failure because of the mangled hands.
I considered this try a failure because of everything.
That first one was just the name. Then I tried throwing the entire text at it:
I spent an inordinate amount of effort of trying to get these two stupid joke images. For the one with the girl the problem was trying to get all of the elements in place (it especially wanted the genie to come out of the top of the lamp instead of the spout, for example). For the one with the guy it didn’t want to make it at all: in three four-image sets on Sora each had three “failed” images. Then both of them needed some outpainting to get decent framing.
On a positive note, I’ve discovered that Night Cafe’s new outpainting option is smarter/better than any that I’ve used in the past. Cost 1 credit to outpaint an image. (Which isn’t a big deal for me because I religiously collect the free daily credits for my two accounts but never use them and have over 2,000 credits accumulated.)
Oh, and Night Cafe wouldn’t let me upload the guy version for “child safety” reasons. But making the image into a negative worked for uploading and outpainting.
Speaking of pet groomers:
Fooling around making parody album covers featuring Dr. Teeth and the Electric Mayhem from The Muppets. The first two were just word prompts, the last three I gave it the original album cover to reference and it seemed to do much better.
Those are awesome.
While doing character combos, I vaguely remembered some character that ate everything. Some googling proved it to be “Big Mean Carl” from the Muppets. But ChatGPT didn’t seem to know how to make him without a reference image, at least in the one try I made before providing one.
First try was an unambiguous (but hideous) Bart Simpson but Big Mean Carl was a dog. Plus the framing was too crowded. Widescreen tries looked less Barty, it didn’t make Bart small enough, Bart looked in the wrong spot, and the shorts and the eating there-of never came out right, so overall I didn’t get a satisfying image. But Big Mean Carl looked good once I gave a reference.
Sora appears to have upped free images a day from 6 to 12. Add to that the increase from only 1 variation allowed to 4, and in the few weeks I’ve been using it Sora has gone from 6 free images per day to 48.
I know the punchline for the first joke:
“You think I asked for a 12 inch pianist?”
…but I’m drawing a blank on the second one. Little help?
“Hey, I wanted perfect tits and ass!”
Ah, so that genie wasn’t hard of hearing, just doing the classic djinn fakeout wish-- granting the letter of the wish but not the spirit of it. Don’t think I ever heard that one before 
Haven’t tried one lately with the new image generator, but Dall-E, without guidance, almost inevitably drew blue genies looking like the Will Smith version from Aladdin.
In the prompt for the earlier Alice/Dorothy image, I replaced “35 mm disposable camera selfie” with “Arthur Rackham painting”.
Replaced that with “A scrapbook-style scene made entirely from layered paper cutouts and textures.”
(Both ideas found in prompts on the Sora feed.)
My “Inappropriate Little Golden Books” series.
I wasn’t clear on that. I meant, if I asked for a genie, it would almost always make them blue Will Smith versions.
Hilarious!
Even more hilarious!
Turns out I didn’t have to give it the text. It knew what the parts were called anyway.
