Whiledさんのイラストまとめ

6/ Hero.

This is another one where I cheated, because you see, Hero's hair is non-conformist, it's not easily described to an AI to replicate, so I used img-2-img again.

#OMORIFANART #AIArt #stablediffusionart

5/ Kel.

I did cheat with this one, overall text-to-img prompts are... spotty, so why do that when you can tweak an existing image to your liking! T-shirt designs stay more consistent that way.

#OMORIFANART #AIArt #stablediffusionart

4 / Basil.

The first one I started with, Basil's design lends itself well to be generalized, but the AI really likes turning Basil into a girl if not given the proper weights.

#OMORIFANART #AIArt #stablediffusionart

3/ Aubrey.

It's a thin line between getting absolute sexy and cute, the models don't like something that doesn't conform to either. Also, that hairband was a pain to get right.

#OMORIFANART #AIArt #stablediffusionart

2/ Sunny.

He was the most generic of all, amusingly the tag Omori has weight upon the models used hereforth.

#OMORIFANART #AIArt #stablediffusionart

#AiArt #BreathoftheWild
When attempting to generate more than two people, the AI struggles to figure out what details to match to who. Add a background too? Forget it.

#AiArt #Samus
An older generation, though I find hyper realism to be very appealing, it makes very perfect faces which might not be always ideal.

#OMORIFANART #AiArt
The previous version might have been a little unwholesome, so here's a new one.

#OMORIFANART #AiArt
I'm not as happy with this one, the more detailed the more need for a composite of multiple images which lowers it's quality overall.

#OMORIFANART #AiArt
More AI generation, using an image for input.
Can you guess what image I used for reference?