Inspired by a recent NAI v3 gen of Anon on 4chan and after failing miserably to replicate it in Pony (probably a skill issue) I decided that some cooking had to be done. I didn't use realistic images in the dataset but in my tests it seems to still work on the Real Pony Checkpoint.<h2 id="heading-1052">Info for v1_male_on_female</h2>Although v7 has the data to generate the male on female piggyback concept the probability is low. This version has only images with a male sitting on the back of a female. I was able to find only 16 decent images for the dataset so this LoRA isn't very flexible but may still be usable. It also has difficulties generating good eyes.<h2 id="heading-1053">Info for v7_female_on_*</h2>Example Triggers:<ul><li>piggyback riding, sitting on person, all fours, femdom, girl on top, human chair, 1boy, 1girl, from side, profile</li><li>holding leash, bdsm, gagged</li></ul>Just sitting:<ul><li>human chair, sitting on person, all fours, girl on top, 1boy, 1girl<ul><li>crossed legs</li></ul></li></ul>Trained perspectives (some work better than others):<ul><li>from side</li><li>straight-on</li><li>from behind</li><li>from above</li><li>from below</li></ul>Trained but lower probability to trigger:<ul><li>boy on top</li></ul>

Files with SHA256: 71a5d8b78cf1336b3ddcb174ca27067f2d60f2aef28c3643b019f1d772e81f9b