Between these two experiments, including a few dozen other generations that remain unposted, one thing I can say for sure is that for living subjects, it's a great way to get the kind of anatomical wonk that older models are (in)famous for - and it makes sense why, the model is trying to make something that looks like a certain subject...but once it starts to look too much like it, well, shit, we told it NOT to do that! Break something up! Given that I love that kind of wonk, I think I've found a useful tool for myself.
One more living subject, and let's get even more abstract with our direction here:
I then tried to see if, learning from the animal subjects, I could make it more likely to return one of my favorite "mistakes" - making it impossible to discern the point where a water area ends and a sky area begins. I wasn't immediately successful, but I came up with some results I found pleasing regardless-
Positive prompt: Secret hideout in a cave behind a waterfall in the foggy forest on a floating sky island in fluffy clouds
Negative prompt: hideout, camp, campsite, home, abode, house, dwelling, rest, shelter, waterfall, water, cave, grotto, forest, woods, woodland, trees, fountain, cascade, pond, stream, lake, river, brook, puddle, creek, pool, beach, ocean, sea, cloud, clouds, sky, cumulus, cirrus, nimbus, fog, storm, rain, sunshower, falls
It seems that with landscapes it's got a much clearer and more specific "idea" of what a [SUBJECT] without [SUBJECT] looks like; it's more inclined to invent very specific, very consistent unasked for related elements. With the animals, I was tweaking the weight on the positive prompt to avoid getting straightforwardly just what I had positive (and negative) prompted, but with landscapes, I just get... almost something else entirely.
So how about inanimate objects? Let's try a ship, perhaps?
Positive prompt: A huge sailing ship with brilliant prismatic crystal sails on a stormy, turbulent sea of sunset clouds
...okay, I'm in love with the un-ship. It truly does manage to consistently give me results that look like, yet entirely unlike, a ship. It is everything I love about AI as a medium. More than that, it is my friend.
At lower positive prompt weights, they only get even more beautifully chaotic.
I want to live on one of these (in an alternate universe where they're geometrically possible and structurally sound, that is).
Failing that, I will be featuring them a lot from now on.
All images generated using Simple Stable, under the Code of Ethics of Are We Art Yet?
Yes I do think that the ban on porn in America is directly tied to pushing people to use generative ai because wdym you don't want kids viewing potentially harmful shit yet you push for them to use ai that could make littleraly anything.
I also think it's to scare people from posting online because you can take a screenshot of anyone's face and tell AI to just put it into a porno and potentially ruin their entire life, which when put into the circumstances of today are very likely with how divided people are nowadays with them going to extreme lengths to ruin someone's existence just for opinions or simply just being something that you can't control (skin tone sexuality etc).
It would also benefit anyone who has invested in gen ai like anyone from the upper class.
