This is my Ass On Glass LoRA. Frankly, I'm starting to run out of ideas. If you have any idea that you really want a LoRA for, I'm open to ideas. This took two attempts to train correctly. My first dataset was not up to standard, and my second was with the caveat that it had more furry content than I would personally permit, so I would put '(furry), (animal ears)' in your negative prompt. A strength of 0.6 to 0.8 is good. I use 0.75. I will admit, to have the full glass-squish appearance, it might require a higher weight. This prompt is still tricky with shots from below where the character is sitting on a glass table. The keyword is "ass on glass".

Files with SHA256: fdd061d14c5ff6988f255fb5ede482ddf0f94ac28ba35068d4bf96993742d9e3