本 LoRA 专门针对《鬼灭之刃》中的蝴蝶忍进行身材比例强化。在保持角色标志性服饰和面部特征的同时，重点强化了丰满的胯部、肥臀以及曲线美感。模型基于 Illustrious-XL 训练，具有极佳的画质表现和人体结构稳定性。<img />Model Details / 模型规格Base Model: Illustrious-XL (v0.1)Dataset: 69 High-Quality hand-picked images + Flip Augmentation (138 total training steps per epoch).Network Rank (Dim): 64Network Alpha: 32Size: ~435MBTraining Style: Focused on anime fidelity with specific body-type bias.Trigger Words / 触发词Base Trigger: shinobu_kocho, kimetsu_no_yaiba, 1girl, solo, back_view, from_behind, low_angle, ass_focus, ass_shot, huge_ass, wide_hips, thick_thighs, plump_thighs, bottom_heavy, curvy_figure, standing, legs_together, arms_at_sides, arms_outstretched, hands_outstretched, haori, white_haori, black_pants, tight_pants, thigh_gap, purple_hair, hair_bun, butterfly_hair_ornament, hair_ornament / 使用建议Recommended Weight (推荐权重): 0.8 - 1.0Sampler (采样器): Euler a / Restart / DPM++ 2M SDESteps: 25-35CFG Scale: 5.0 - 7.0Resolution: 896x1152, 1024x1024, or higher.Positive Prompt (推荐提示词):shinobu kocho, kimetsu no yaiba, demon slayer uniform, butterfly haori, haori, hakama, purple pants, wide pants, puffy pants, voluminous clothing, wide hips, thick thighs, exaggerated proportions, black hair, purple hair, gradient hair, short hair, butterfly hair ornament, purple eyes, holding sword, katana, weapon, both hands on sword, combat stance, crouching, squatting, spreading legs, ready to fight, solo, simple background, gradient background, anime styleTechnical Note / 技术说明The model uses a Rank 64 / Alpha 32 architecture, which strikes a perfect balance between learning capacity and model stability. Thanks to the flipped data augmentation, the model handles both left and right-facing poses perfectly without "pose-locking."本模型采用 Rank 64 / Alpha 32 架构，在参数量和泛化性之间取得了完美平衡。由于加入了镜像反转增强，模型可以完美处理各种左右构图，不会出现动作锁死的情况。Disclaimer / 免责声明This model is for artistic and creative use only. Please do not use for any commercial purposes without authorization.本模型仅供艺术创作参考，请勿在未经授权的情况下用于商业用途。

鬼灭之刃-蝴蝶忍（丰腴/肥臀增强版）- 基于 Illustrious-XL

对后背签名和后背视角服装进行优化

Files with SHA256: ff7caea85bef69384e7e2b04d664dbf02e460e61a587cb77c0cc7313545f0598