Not A Play Area (Diaper Dimension) (Patreon)

I probably spent far too long on this one, but currently getting the AI model to handle coherency across higher resolution images is the next frontier I'm trying to conquer.

It has an internal 'attention' mechanism, which is the pseudo-intelligence which lets it figure out which parts of the image are associated with which other parts, which was trained at 512x512 and can stretch a bit beyond that, but doesn't work so well at higher resolutions. With the new hardware I got this month it should be possible to try training on a newer 768x768 model this month.