suzuran_(hongshan)|铃兰_(红山动物园)：铃兰，Arknights X 红山动物园的联动皮肤。
简介不会分开两个版本写就不写训练底模了，反正训练时候没用自然语言。
• 触发词 (Trigger Word)：suzuran_(hongshan)
• 基础外貌 (Appearance)：1toddler, 1girl, solo, fox ears, multiple tails, blonde hair, white inner hair, twin braids, green eyes
• 核心服装/道具 (Clothing & Props)：aqua hair bow, yellow collared shirt, short sleeves, orange gloves, black belt, white waist apron, teal skirt, purple underskirt, purple leg warmers, black shoes, crossbody bag, holding staff, lantern, butterfly ornament
• 自然语言支持：未原生训练。
若有则展示自然语言提示词：无。
• 推荐 (Recommended)：suzuran_(hongshan), 1toddler, 1girl, solo, fox ears, multiple tails, blonde hair, white inner hair, twin braids, green eyes, aqua hair bow, yellow collared shirt, short sleeves, orange gloves, black belt, white waist apron, teal skirt, purple underskirt, purple leg warmers, black shoes, crossbody bag, holding staff, lantern, butterfly ornament. A full-body three-quarter front view of the character standing on one leg, smiling, holding a staff with a lantern and butterfly ornament, against a white background.
• [0.60 ~ 0.70]：动作与场景自由度高，但细节特征可能会略微下降。
• [0.75 ~ 0.85]：高度还原设定，同时保持良好的构图泛用性。
• [0.90 ~ 1.00]：易出现过拟合。
• 采样器与步数 (Sampler & Steps)：Euler a / DPM++ 2M Karras，28步
• CFG Scale：5
• 道具处理 (Props Handling)：
crossbody bag 为可穿戴配件。holding staff、lantern、butterfly ornament 为同组道具设定；不需要时，请从正向提示词中一并移除。训练集中没有独立 staff tag，建议使用 holding staff。若不希望出现尾巴，请移除 multiple tails，并将 tail 加入负面 tag。
• 训练集：利用GPT-Image-2 与NAI 扩充训练集。
通用负面：watermark, watermark in lower left, text, signature, logo, lowres, worst quality, low quality, normal quality, jpeg artifacts, bad anatomy, bad hands, extra fingers, missing fingers, fused fingers, blurry face, deformed face, ugly, poorly drawn, 3D, realistic, photorealistic
