🪨 (@Varpie)

@nixCraft This is a pretty complex problem, and I think there are 2 main things to consider: the ethical aspect of using GANs and the integration in your workflow (which is true for artists but also for coders with GH Copilot or copying code from ChatGPT for instance)

On the ethical side, GANs may be trained on copyrighted material, which means something that is AI generated may not be usable for commercial projects, and in a way you are "stealing" someone's work if that is the case. Which makes using them more difficult if the objective is to just become a prompt engineer and let the AI do all the work for you. And it wouldn't exist without the artist, and isn't able to generate new styles either: something it isn't trained on will not perform well. Since each artist has their own style, unless you fine tune it for your style, it won't look like what you want, but like someone else's art.

Which brings it to the second point: it is a tool in the end, just like digital painting. Some artists enjoy the more organic feeling of using a pen and paper or painting on a canvas, but digital painting is more "efficient" because you can more easily color large areas in a pixel perfect manner, for instance. I believe that as a tool, it could be used to increase the artists' productivity a lot: it won't give exactly what the artists have in mind, but if they can draw the base layer, let the AI fill in most of the details and complete it with their own vision, it can be a lot faster without having to lose the artists' identity.