We are not going to use any gen AI for future projects, so my plan was to keep working on new games and leave this one as it is. But it's a fair point about this normalizing gen AI. I mean, I don't think the game is big enough for that to be a problem, but we are quite against how the technology is used nowadays so it's hypocritical to keep those at least.
So I asked a friend who is an artist to redo the frescoes images at the end of the game. At some point during the next weeks we're going to update the game with those.
We also have some 'vision' videos, but I can't really think of a replacement for those, so we're not changing them at least for now. As opposed to the frescoes that were generated using Midjourney/Stable Diffusion, those were generated using VQGAN+CLIP. This one wasn't really a mainstream tool afaik, as their data models are really rough. And we selected a model that was suitable for commercial purposes.
So I'd say VQGAN+CLIP doesn't have the problem of normalizing the technology or harming the environment, as datacenters weren't created for the sole purpose of running it. The dataset we chose can be used commercially, so I assume it was trained with images without copyright. I couldn't find the details of how it was trained though, only that it was trained with Flickr images. Some of those images are allowed for commercial use, so I think it's a fair assumption that they only chose images without copyright.
After we remove the gen AI frescoes hopefully you can try the game!