I hope StyleGAN isn't too tiresome here by now. This is a Twitter bot I put together in the last couple days. The model still needs to train longer, as it hasn't converged, but I might not finish it so I figured I'd put up what I have so far.

The GAN is trained on a database of mugshots. The captions are a mix of curated completions from GPT-J, variations written by myself, and jokes written entirely by myself; all arranged in a phrasal template. The bot is meant to make you chuckle slightly while drawing attention to 1) the ethical and privacy issues inherent in such datasets; and 2) the obscene number of human beings packed into cages by the racist US "justice" system.

While it certainly has room for improvement, I hope this is an acceptable first post here and that someone finds something of value in the project. I fully expect this to sink into obscurity, but in case it doesn't, feel free to criticize me, ask questions, etc.