Unfortunately, the "wonderful" (sarcasm) CivitAI flagged and removed two images—one in a qipao and one in a bikini—as "R-rated." Just know that this LoRA can generate her not only in her default outfit but potentially in virtually any outfit at all. (And no, I will never, ever share my prompts.) After finishing the game completely, I realized that my second-favorite character is none other than Anzu. Yes, she has very few actions, dialogues, and so on. But I really liked her appearance and the personality I managed to grasp during... our rather brief acquaintance with the character :(I decided to create a LoRA for her, but there were even fewer images of her than there were of Kanna Kizuchi (my previous LoRA). So few, in fact, that the dataset consists of only 16 images, 90% of which are just her in-game sprites :(Character prompt: AnzuCute, short hair, forehead, two-tone hair, colored bangs, red bangs, blue hair, medium breasts, red eyesClothing prompt: two-tone skirt, red skirt, aqua skirt, miniskirt, orange sweater, ribbed sweater, turtleneck, red socks, kneehighs, long sleeves, detached hoodThe following issues have been observed with this LoRA:1. The LoRA has overfitted heavily to her clothing, so prompts like "hood down," etc., will still generate her hood.2. Unlike the Kanna LoRA, this LoRA has absolutely no knowledge of the collars from the game.3. I did not train the LoRA to generate her clown mask.4. I trained for 20 epochs and, during testing, found that this epoch (14) shows the fewest hair-color issues. Across more than 200 pictures, I haven't encountered a single incorrect hair-color placement, though I still can't completely rule out the possibility of errors. 5. Her standard footwear as part of her "clown" outfit is not learned by this LoRA (and I didn't aim to achieve that either).This LoRA is from epoch 14 out of 20, so it might be slightly overtrained. If comments confirm my concerns, I can upload an earlier epoch version.

Files with SHA256: 3216ff98cd441e4f7819357a12ca00250228afd95ff6f2f9d58f196009a39110