I figured this day would come. Honestly I thought it would be faster after Dalle 3 back in 2023. The world of image generation has been largely driven by enthusiasts on the internet rather than companies due to their rampant filter and copyright restrictions. Open Source was going to win eventually though, and here we are now.
We get a lot of AI art submissions in the submit box, and 90% of them are completely unlabeled. As of the last month or so, the solution has quickly shifted from catching typical AI errors, to me running around learning what LoRAs and models people are releasing while being wary of new artist styles and keeping an eye on existing artist galleries to cross reference submissions that look like theirs.
I'll be doing a longer editorial once OpenAI drops their new image generation later soon™, but I figured I'd get a general update out now. So have a mini writeup update thing! Continue below.