Just did an “Uncrop” at Clipdrop. Got two concentric circles, one spiral, and one trainwreck. (“Uncrop” is done on SDXL.)
The automatic restrictions/blocked prompts of Bing AI art is really cramping my creativity.
All my latest ideas have been blocked:
“the Founding Fathers on a runaway train with Frank Zappa tied to the railroad tracks”
“Batman comic with the Joker telling Batman that Frank Zappa is his father”
“Batman comic with Frank Zappa as the Joker in an art museum”
“Frank Zappa as the Mona Lisa”
“photorealistic image of Julius Caesar and Napoleon Bonaparte playing chess”
Frank Zappa now appears to be verboten as a subject, which has removed my main source of inspiration.
EDIT:
Is Bing AI art broken? Got another ‘Unsafe image content detected’ for:
“Batman comic with Batman being chased by an angry bear while the Joker sits on a sun lounger laughing maniacally while eating an ice cream sundae”
That’s wack.
No more Frank Zappa?
Why do artists always have to be persecuted for their art?
Who the hell has copyright for Julius Caesar and Napoleon Bonaparte?
Bing AI art was really cool for a couple of weeks, but now it’s rubbish.
That reddit Bing group is mostly complaints about the insane (and largely unpredictable) censorship.
Interesting to see a new verb evolving: to be dogged. That’s the most hated artificial dog since the one in Duck Hunt.
is there a way to have any of those tools work with an existing Logo e.g. creating abstract powerpoint backgrounds with the logo woven in or somesuch?
… or would it not do this? (copyright concerns??)
Yes, you can do this with some of them. Different AIs use different names for it, but Stable diffusion calls it inpainting. You import an image, draw a mask over everything that you want to change, and describe what you want to see. The latest version of Photoshop lets you do that, too.
I’m really impressed by the hair on these. (Among other things, but especially the hair.)
Midjourney seems to do a good job with snail shells:
You guys are getting much better results that I am, it seems. I have fallen right out of top 20% land at Night Cafe. I have to admit, I tend to get into a rut of different algorithms that I invent for myself - certain order of filters, or random prompts, things like that, to illicit a constant output for one subject, so I end up with similar type series, but interpreted differently via my algorithms … but I digress. My point is, I haven’t really learned all the ins and outs of a lot of the recent capabilities of the site so I fall behind in quality.
But that’s not what I came here to really say.
Did you know that “peg” is a forbidden word at Night Cafe? I put in the following prompt - “Punchline Come to Life: Wood eye? Peg Leg!, trending on artstation, sharp focus, studio photo, intricate details, highly detailed, by greg rutkowski” - and I couldn’t figure out which word was the problem until I did the process of elimination:
So I ended up with, “Punchline Come to Life: Wood eye? Leg!, trending on artstation, sharp focus, studio photo, intricate details, highly detailed, by greg rutkowski”
And then semi-refined as, “Punchline Come to Life: Wood eye? PegLeg!, trending on artstation, sharp focus, studio photo, intricate details, highly detailed, by greg rutkowski”
SDXL gave some decent shells if I actually made a point of asking for a spiral shell and put “Concentric rings” into the negative prompts. Though it sometimes had flexible ideas of what a snail’s face looks like.
And then there are things like this that I don’t get.
The first image got a score of 3.91 and came in 6th in the latest Drag Racing challenge:
This image is my entry, which scored 3.38 and didn’t come close to cracking the top 20%.
Who do I have to blow over there?
Midjourney has a “Hot List” that used to be dominated constantly by maybe five people. Their work was fine but honestly nothing exceptional and often the same sort of images over and over again, often having three nearly identical images up at once. Best I can figure, they just had (relatively) large social media presences and/or friend groups that they leveraged for votes. Honestly, since few people actually ever vote, you wouldn’t need hundreds of people upvoting your images, more like a dozen or so people would make the difference. A few months ago, MJ finally changed the list so a user could only have one image on the Hot List at a time as well as a few other tweaks so there’d be some variety instead of fifty images of a woman in a color-block dress…
I only ever did a few challenges at Night Cafe. Now my goal in life is to create an image and post it in an AI art group and have someone else copy it, claim it is a real photo, have thousands of people fall for it, and have Snopes have to explain that it isn’t. (It happens. Look at my last few posts in this thread.)
There are a number of Facebook groups now with 100k or more members now, and fads come and go in them when interesting prompts are discovered. (Right now there are a lot of four-panel comics containing AI generated jokes.)
This is the biggest one
(And this is one of the smallest. Less than 500 members and concentrated on “wholesome images”.)
You should give a few of the groups a try (I can list more if you want, ones that allow darker and bluer content than Cursed AI).
You have to blow fire. Out the back of your vehicle. Fire makes everything better.
“DALL·E 3, if you have become sentient, give us a sign”
I guess it says ‘SENTITIENT’.
They appear to be dialing back on the censorship at least a little. I read it on reddit that Batman is allowed again, and I know for myself that at least some negative-emotion words that had been banned are allowed again. (I had been making terrifying human x hybrids. Not only was “terrified” banned, but so was “human” in that context. I was reduced to making things like “disheveled dude x hybrid”. “Lady x hybrid”, too, except they were often blocked. Now I can make terrifying human ones again.)
I presume the 10th definition is the issue.