Sadly, there's nothing for us to do but to complain not only about Perchance but also other corporations. Actually, we must complain not with Perchance, and solely with other corporations (ok, maybe we can complain with Perchance, but just a little), since Perchance is a project that is not profit driven maintained by a single person and it uses the models of these other corporations.

All text and image models are extremely biased. In text for example, I often have a great difficulty because it tries to shove LGBTQ+ in everything. Nothing against LGBTQ+, but when literally all cyberpunk vampire characters I try to generate is said to be LGBTQ+, and when I try to say "cishet woman" and it makes sure to say "she is a cis woman but has no genitals whatsoever", we have a big problem.

You can see other problems in the generators. For example, everyone is always way too sexualized, and some times, problematic. You may try to write as much "flat chested adult" as you want, and most results will still be a huge breasted bimbo, a loli, or a huge breasted loli. Worst of all, their waist is always thinner than their head.

One of the worst offender is when you try to generate a "mixed race" or "ambiguous race" person. All of them are 100% black. "Mixed race" doesn't mean 100%. One can even be mixed between white and Asian.

And the worst case I ever saw was when I asked the text AI to judge me based on my tastes. It said that I liked purity because I liked milk and ladies with light skin. Bro... I do like the idea of "purity", because hygiene is cool, but... First, milk is just a matter of being healthy and tasty. Second, I like light skin because I like the color white, and saying that light skin is anything related to purity is racist as heck. The same AI that throws stones at me for making jokes with races is the same AI that says that light skin is completely related to purity.

Other bias it has is always trying to make anyone it perceives as minority a hero. Once, I had excessive problems in writing a guy who was unredeemable evil without sad story or whatever and who happened to be black, all problems caused by the AI trying to give him a reason to be evil because, apparently, black characters don't have the right to be mustache twirling villains. And other time, I wrote about the story of a guy who was so stupid and his stupidity made everyone powerless, and a minor character was a feminist girl who gave up on fighting and became a drunkard trying to forget ever meeting the stupid guy; the AI kept pestering me to turn the feminist girl into a hero who'd be the only one to properly face the guy's stupidity and even changing the whole course of the story.

Another problem is that EVERY TIME I ask for a prompt of a smut image of a woman and don't mention how she is looking, the AI always come up with "she looks confident and unapologetic". At first I ignored, because confident women are hot, but bro, EVERY TIME? No matter the fricking thing she is doing, "she looks confident and unapologetic". She is posing naughtily: "she looks confident and unapologetic". She is showering: "she looks confident and unapologetic". SHE IS TAKING A DAMN POOP!: "she looks confident and unapologetic". DAMN! Can't women, simply, like, exist? Must they look like a propaganda poster 24/7?

And the worst is when I try to complain about anything. It keeps blaming me, until I write the words "I'm a nigga" or similar minority. I shouldn't need to show my ID card to have a proper engagement with AI or anyone else.

Other weird patterns I noticed on my usage, in this case, specially with image generation:

• Small breasts/flat chest: "she is young, short and has a huge ass", even if I said "tall, skinny, mature".

• MILF: "she is unshaved", even if I said "completely shaved/hairless neck down".

• Hair with bangs: "Blunt bangs", even if I specified another type of bangs.

• Depressed: "She is into alcohol and other drugs", even if I said that they are depressed exactly because of bad experiences with other people using alcohol and other drugs.

• Likes rock: "She is full of tattoos and piercings", even if I said they don't even wear ear rings.

• Is into tech: "Used energy drink cans all around", even if I specified "only milk".

• Is productive: "Cup of coffee in hand", even if I specified "only chocolate".

• Yandere: "Japanese high-school outfit", even if I specified "30-years old Caucasian".

I have a doc of ~48k characters / ~8650 words solely of bad experiences I had with AI, mostly with the text one, but also with the image one, and almost everything thanks to Perchance. I also have 2 other beginnings of medium length stories based on such experiences.

In the end, Perchance is cool and all, but it's far from perfect, and the problem is less with Perchance, and more with the corporations that made the AIs that Perchance uses. There's little hope to solve it (yeah, I'm a pessimistic person). But you can try to complain to other corporations, specially the ones that cause the problems we face.

In your specific case, one could say that it's your simple prompt's fault, but here I am, going beyond it: I tested with better prompts and I still noticed way too much problems, both in lack of variety when it should have, excessive variety when I asked for no variety, and when it takes a single token minority and shoves in 100% of the results and calls it "variety". What we face is not Perchance's fault, it's the industry's fault.