
Sadly, this character has no LoRAs published so far and I needed to remedy that. Unfortunately, there is very little art of her to be found online. Thus, this model was trained from in-game screenshots, which worked very well at capturing her features. Unfortunately, I did not sufficiently vary the locations where I took these screenshots and the LoRA is overtrained on certain features like wooden floors and one-point perspective. It's manageable, but you may need to describe the background in detail to overcome this bias and putting "wooden floor, school hallway" in the negative prompt may help. I recommend keeping the strength at 0.85 or below. The trigger word is "bryuki" and if you need to reinforce her appearance, use "brown hair, brown eyes, hair bun, twin bun" and "brown school uniform" does pretty well at her default outfit, but some other DLC outfits were also in the training data, so it's flexible.
描述:
Old dataset retrained on Pony Diffusion v6.
训练词语: bryuki
名称: bryuki_v3_PONY.safetensors
大小 (KB): 223104
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success