Oh, I dunno. I get inspired when AI dungeon gets stuck in a horror loop. 
@erinrae26381 if the “all-characters-look-same-y” service is the one I’m thinking of, I don’t think it is actually generative, it just blends two (user-selected) pictures together… but maybe we’re thinking of a different one.
It’s been a year or two since I was on that site, so I think you know more than me - I’ll take your word for it!
I mean, I apparently tried it out in 2020, so it could also have changed. (I wanted to make a portrait for my Sidestep and had no idea how they looked like, so I figured why not. Kinda liked how it turned out, too. Then I traced it to get a feeling of the shape.)
Also worth noting for people who are sure they can always recognize AI writing—there are different generative text AIs out there, and they don’t all have the same style. If you’re good at recognizing ChatGPT—which does have its own catchphrases and tendencies—that doesn’t necessarily translate to recognizing other ones.
Which applies to other things people describe; they don’t all have the same approach to what material is restricted or not, for example.
It’s all quite complex.
That is certainly true for the image generators as well (although you’re more likely to detect AI-generated image from having floating body parts than you are to detect AI-generated text from incomplete sentences).
There’s certainly certain characteristics in certain generators you can detect in certain conditions.
You should check out online resources like https://theeverybodyproject.com/ or street style photographers like Johnny Cirillo (@watchingnewyork) • Instagram photos and videos or even go on royalty-free image sites like pexels.
It’s totally fine to use any reference you like if it’s for personal use like learning to draw. You could trace pictures of Spiderman and Kamala Harris all day long if you wanted to. But if you want to publish your drawings publicly, especially for any kind of profit, it’s better if you get permission from the subject and take your own reference photos. Street style photographers usually just say something like “Hi, I’m ____, can I take your picture?” or “Hi, I’m ____, I’m a photographer, and I snapped this picture of you and your friend. Is it okay if I use it in a project I’m working on?” They might ask for a few details like where and how it’ll be used, and they might say no, but they might say yes!
I’d say that depends on what kind of reference picture it is (I’m not going to go into the wild to take photos of emus or Smilodons to get a sense of how they look like, thank you very much) but it’s not like I disagree.
(Also 3D mannequins are great, but they don’t always look properly human, which is sad, since I prefer those.)
Your cute Dog-Pegasus is a new nightmare creature for me. 
Ha, sorry
Haha, I was gonna say that it walked that line for me as well.
@LiliArch that’s an excellent Sidestep face: 11/10 perfect, haunted thousand yard stare. I’ve never actually drawn my Sidesteps - now I kind of want to lol.
Hmnh, I really need to do more practice… I don’t know if I’m just rusty, or if it’s drawing digitally instead of on paper that’s throwing me off, (probably both,) but I don’t seem to be able to draw a proper jaw anymore.
Well, at least I still know roughly how human limbs behave, which is more than I can say about any of the AI generators I’ve tried (still like making moodboards with them though).
Oh, I know I’m well out of practice lol. There are a handful of free/cheap live-model drawing events around me and I should probably haul myself there once in a while.
I’ve never really been as comfortable with digital media as with physical, so I feel you there. Pen, paper, and watercolor has always treated me best.
What’s the beef with AI? And I’m pretty sure that CoG company will allow AI in your work if you only use it to a certain capacity.
Like Grammerly for instance, one could write out a sentence, then use the program to enchane it for better engagement for the readers.
And besides the point, if a story is good regardless, why would you refrain from enjoying it just because its AI generated?
)They will not allow anything with either AI-generated text or images to be published, which is something that the moderators have stated more than once-- and if you try to be sneaky, they will pull your game. They’ve already done it to one author who tried to be sneaky about their AI usage, so they are not playing around-- which I am extremely glad to see.
To me, there is no such thing as “enhancing a sentence” via AI. You can either edit your own writing, or you can generate complete garbage with AI; doing the latter is not an “enhancement” of any sort.
Using a spell-checker is not AI, nor is it “enhancement.” It’s when you use a damn program to generate your writing for you that I find completely unacceptable.
…I’m suddenly having an urge to pull some statistical data and build a randomizer based on that.
I’m trying to understand what benefit this has over regular procedural generation, but all I can think of is if it’s a geographical map and the AI was specifically taught to create realistic geographical features (in which case it was probably trained by real world maps, which… I’d assume isn’t stealing, because who owns rights to Earth’s geography?).
Given how random geography can be. I find map making through A.I to actually be fine, paticularly for projects that are small or minor. Such as one shot DnD or small books where a map is not necessarily a major aspect but can help aid in viewing the situation.
But for maps that take a big point in the plot. I.E Maps importsnt enough to span multiple books or used as plot devices through its geogrpahy. Its better to try and hand make it. With A.I only used to provide ideas through interesting geography it generates.
I was mostly thinking using AI in things like making sure the rivers flow in correct direction and the like (which I’d assume an AI trained with correct geography should be able to do, but I also haven’t seen any, so I can’t be sure). Like a grammar checker for maps.
(If you just want to make a random map, you can do it with way less computing power procedurally without a generative AI.)
…why would you need to draw it by hand instead of using, say, Inkarnate, if you’re not confident in your map-drawing skills? 
No Grammerly? Seriously?
The fun thing about neural nets (which is what LLM “AI” really are under the hood) is, they’re pretty common solution to control behavior of artificial opponents in all sorts of games, and have been around for much longer than the current hype. There’s likely quite a few games you’ve enjoyed thanks to the very thing you “universally hate”, without even realizing it.
And here I thought NPC AIs are just glorified decision trees without any kind of machine learning (the ones where adapting to player’s strategies is a big thing of gameplay tend to advertise it, by my experience).
