Trying to get the perfect image is an exercise in frustration. The AI models are not good enough yet, and you will always randomly find things that aren’t exactly what you want.
Now add something absurd!
The future needs more high-tech corporate geese
Sweetie - she was joking
“Perfect” as in “meeting the bare minimum instructed”… then failing 62 times… and the 63rd being OK’ed out of tiredness. Making specific looking people react in specific ways while wearing specific clothes in a generic environment shouldn’t be that difficult but it is… even the part with the jerseys looking like they’re worn by women, not men, with the corresponding curves.
The issue is that the ‘AI’ is not actually thinking - what it is doing is generating a probabilistic version which inevitably drifts away from what you asked. It can generate some interesting stuff, by random chance, but it will never really create what you want.
Yes, I know that’s how it works, they’re just stochastic speculators, but even then I kind of expected that nothing I asked was too unusual and the model had a good training with it. I will keep things simpler in the future, the more elements in a prompt, the higher the chances that AI fails even when the prompts are improved by AI in order to be easier to manage by AI…
It is impossible with humans too, just using words alone. Imagine people have their own ideas how something should look, especially when not described in detail. Some even think their ideas are better than what was provided to them.
Maybe we are all just a probability, maybe we are already a solved probability also. I feel like it is solved, honestly.
My ideas are the best!
After trying to add another of my chars to the “detailed prompting” and seeing Nano Banana fail miserably to capture her complexion and do a iffy job about most of her face features, I’ve decided that getting AI to make my chars look like themselves is above my effort range.
I don’t need those images that bad, and I’ve not even tried getting an AI to do something really original like my own imagery.
AI is a niche tool, with a very small and essentially worthless niche to fit in.
I hate AI. Does AI hate me?
Yes, Ish is not holding a fan of cards. Guess who forgot to rewrite that instruction…
Seriously, why are oversized comic characters so much easier to do than real ones…?
I’ll need time in my thinking corner.
Take your time.
Look at how hard even today it is to learn this game. I hate AI only because I watch a lot of history, but AI is never as accurate as a human dissecting it. AI in that context generalizes it and does not research back far enough, hence not making as accurate as it could be.
AI art is a different story. Look at what Zaera does with art, always working to improve. Learning small new tricks along the way. Look at what Aiko does with music. Some of which to me anyway, it should be background music in some scenarios in the game itself!
I busted your butt in the past, but you explained to me so well what you were trying to do that now all I can do is try to encourage you to continue.
Take your time, just don’t give up!
They are in a small room.
My guess is that when generating the comic it defaulted to a symmetric composition of the walls and characters. Although hotel rooms in Venetian palazzos aren’t known for their dimensions, they usually have walking room around the beds… But the characters look good and Yiole is properly towering above the Fazmarai cousins (160 vs 165 vs 196 cm), something the realistic images failed to do.
Now I’ll be thinking about using comic as a medium for images…







