Advice on Using Generative AI to iterate on art assets

[DrDankstone]

Hi-

Long time listener, first time caller

I am an artist who has attempted several times to make my own games. I always hit a wall generating all of the poses, expressions, etc. for all characters because it's so time consuming.

I'm wondering how those who appear to be using AI to create their art assets are coming up with such consistent yet specific images?

I've played around with prompting and the like, and can get good results. However, I'm not interested in fully generating assets. I'm wondering if there is software or a proven workflow that would allow me to do something like:
-take a hand-painted model sheet, and apply poses to that character.
-train an AI to reproduce my specific illustration style.
-apply a prop or costume to different body types automatically
-Generate a series of pre-defined expressions for a character

All story, programming, and original art would be created by me. I'm just looking to see how I can automate some of the iteration needed.

Thank you in advance for any advice!

 

[red_faced]

If you have enough of your art already, you could go down the path of training a LoRA for your art style (and for the individual characters). If you want to see what that would look like, the Summertime Saga Fanart thread has a lot of AI art that people made using LoRAs trained from the art in the game.

It's quite good at getting images that look like the characters in different poses, but it struggles with having multiple characters in the same scene without "blending" them together.

 

[DrDankstone]

If you have enough of your art already, you could go down the path of training a LoRA for your art style (and for the individual characters). If you want to see what that would look like, the Summertime Saga Fanart thread has a lot of AI art that people made using LoRAs trained from the art in the game.

It's quite good at getting images that look like the characters in different poses, but it struggles with having multiple characters in the same scene without "blending" them together.

Thanks for the reply, that's helpful!

I think 70% of the time I'd be generating solitary character assets on alpha, but definitely want to have the characters "interact" with each other as well