LoRA for a permanent metal cage, trained for bara images (human and furry).

I used a large selection of my own generated images, and labelled them with a wide variety of tags listed below. Many of the tags give decently consistent results, but some of them aren‘t as solid.

V2 NOTE: The newer version is vastly more reliable, just add ([tag]:1.2) when some specific tags aren't reacting the way they should.

V3: There were some other things in V2 that were bleeding into the model that I wasn't happy with, so I redid this one with a more strict but larger dataset. Should now be a lot less style-influential, and the wanted features should show up more reliably in the results

Trigger V1: permanent metal cage

V2: permanent_c4ge

V3: permcage

Other tags: from front / three quarter angle, arms behind back / arms up, vertical pose / leaned back / leaned forward, feet up, on floor / suspended in air / wall-mounted