Trained on 176 images of facehuggers on 1girl, either on face, ass, or pussy (or a combination). A bit overcooked, I do notice some prompt adherence degradation, but your experience may vary.It sometimes makes skin brown for some reason (weighted dark skin too high in my slop dataset then forgot to tag it). If you don't want it you should just add dark skin to negative prompt / NegPiP.PromptingPretty simple, follow some variation of <code>a facehugger is [mounted/attached] to her [face/ass/pussy].</code> Examples:<ul><li>one on face: <code>a facehugger is mounted to her face, its tail coiled around her neck</code> - helps with tail, doesn't work on other parts</li><li>one on pussy: <code>a facehugger is attached to her pussy</code> - tail prompts don't really work with non-face facehuggers</li><li>one on ass: <code>a facehugger is attached to her ass</code></li><li>one on face, one on pussy: <code>a facehugger is mounted to her face, its tail coiled around her neck. another facehugger is attached to her pussy.</code></li><li>one on face, one on ass: <code>a facehugger is mounted to her face, its tail coiled around her neck. another facehugger is attached to her ass.</code></li></ul>

Facehuggers for Anima

176 images, batch size 8, 200 epochs, 2e-5 LR. Purely AI slop generated dataset (that's why the LoRA is bad)

Files with SHA256: 329544b16b1dcb6ebe535442bd996fd7af84dba24a7cfe4c271b258ca56d1ae4