Looking for recommendations on ai art generators

[Xymma]

I know there is a big stigma on AI art but I'm done with sitting on my thumbs on my concepts. The trouble is I don't know which program to utilize. I played around a bit with Stable Diffusion roughly 2 years ago but I don't remember it well and there are possibly better options by now. I'm a beginner but I'm prepared to "jump in the deep end" so to speak, when it comes to the program. Quality is what matters to me not ease of use. I'm willing to attempt multiple programs till I find one that feels right for me, I just need some good ones to start me off. I'm sure most can handle either 2d or 3d format which works for me as I haven't decided which I'm going with yet. I like both the art style of Exilium and My New Girlfriend but I couldn't find any info on either thread on what program was used to generate the art. I will say what I have planned is far from realistic in terms of content (And therefor imagery). So whatever I go for has to be able to handle extensive and unorthodox descriptions. (Uncertain if that even matters tbh). Any help would be appreciated.

 

[UnknownManX]

Stable Diffusion is still awesome for art generation. If you'll pair it with Photoshop - you can do anything.

 

[Nadekai]

If you're absolute beginner, I can recommend you

You must be registered to see the links

 

[Satori6]

Quality is what matters to me not ease of use.

That comes down to your hardware, really. That's the big filter when it comes to AI generation.

Using Exilium as an example, I uninstalled it like 20 minutes in, because it was pretty bad, but I clearly remember that there's a girl with 6-7 fingers, and you can see the ribcages of half of them, but they look high quality simply because they're done in a computer that can handle those resolutions and number of steps, even if such simple details like fingers and ribcages were ignored.

I will say what I have planned is far from realistic in terms of content (And therefor imagery). So whatever I go for has to be able to handle extensive and unorthodox descriptions.

The weirder the input, the more you'll need to draw and edit yourself.
Although once again, if you have the hardware for it, you can train your own LORAs to handle weird input, assuming that you have a large enough of samples to feed it.