/ic/ - Artwork/Critique

>>6425156
The ease of anime is a red herring. It’s not that it was easy to replicate, it’s that it is in such high volume, especially with how autistically tagged it was on danbooru, makes it perfect for the ai to train on. Same goes to your typical concept art slop you’ll find in abundance on art station. It’s well tagged and in huge amounts.

The more unique the art style, naturally the less of it there is to train ai on. Not saying it won’t be able to match a close approximation on a small amount of training data, but it won’t be achieving near perfect results like anime because it won’t have all that autistically tagged data in abundance. You can see this in a lot of the trained ai on smaller artists - it will look similar but it’s not near functionally useable like concept art style or anime style bots are.

>Last weekend, Hollie Mengert woke up to an email pointing her to a Reddit thread, the first of several messages from friends and fans, informing the Los Angeles-based illustrator and character designer that she was now an AI model.

>The day before, a Redditor named MysteryInc152 posted on the Stable Diffusion subreddit, “2D illustration Styles are scarce on Stable Diffusion, so I created a DreamBooth model inspired by Hollie Mengert’s work.”

https://waxy.org/2022/11/invasive-diffusion-how-one-unwilling-illustrator-found-herself-turned-into-an-ai-model/

Matrix when?
Left is her art and right are AI generated images.