关于更新1.0：我更换了一个底模，用原来的训练素材重新训练了一遍，现在这个lora在动漫模型中能更好的展示2d效果了，构图也更加好控制了（当然在真实类模型中仍然能正常工作）。但遗憾的是，它对衣服细节的把控力变弱了，即使强度设置为1，仍更容易出现衣服结构上的错误（特别是在和别的lora共同使用时）。所以我没有删除原来的lora，希望您能按照需求下载。About updates v1.0:I’ve retrained this LoRA with the original dataset using a new base model. It now produces better 2D effects in anime-style models and allows for better control over composition (though it still works well with realistic models).Regrettably, it also handles clothing details less accurately—structural errors in garments may occur even at full strength (especially when used alongside other LoRAs). For this reason, I’ve kept the original version available for download, so you can choose the one that best suits your needs._________________________________________________________________________________________________这个lora在真实系与动漫系模型上均有不错的表现，和人物模型融合也不错，不过请不要串联过多lora。触发词为:BFdress，更为详细的提示词：dress,necklace,navel cutout,underboob,thong, panties, high heels,strappy heels, sandals,gold footwear,backless outfit,clothing cutout,强度为0.8，当然，根据需要随时调整。φ(&gt;ω&lt;*)——需要特别注意的是：1.当您使用了提示词中最末尾的backless outfit,clothing cutout时，角色会始终保持在侧面或背面视角并双腿并拢，所以当您不需要这个姿势时，请删除这两个提示词；至于别的服装提示词，请根据自己的需要修改。2.在动漫模型中，衣服有小概率丢失原本的颜色，这时候给提示词中请加上white dress。This LoRA performs well in both realistic and anime-style models, and integrates well with character models. However, avoid using too many LoRAs simultaneously.Trigger word: BFdressPrompt: dress, necklace, navel cutout, underboob, thong, panties, high heels, strappy heels, sandals, gold footwear, backless outfit, clothing cutoutStrength: 0.8 (adjust as needed).φ(&gt;ω&lt;*) Important Notes:1.Using the last two prompts ("backless outfit" &amp; "clothing cutout") forces the character to face sideways/backward toward the viewer with closed legs. Delete these prompts if such composition is unwanted. Feel free to remove/adjust other prompts as you need.2.In anime models, occasional dress color loss may occur. Add "white dress" to stabilize colors.

Cutout dress with high thigh slit

我更换了一个底模，用原来的训练素材重新训练了一遍，现在这个lora在动漫模型中能更好的展示2d效果了，构图也更加好控制了（当然在真实类模型中仍然能正常工作）但是遗憾的是，它对衣服细节的把控力变弱了，即使强度设置为1，仍更容易出现衣服结构上的错误（特别是在和别的lora共同使用时）。所以我没有删除原来的lora，希望您能按照需求下载。I’ve retrained this LoRA with the original dataset using a new base model. It now produces better 2D effects in anime-style models and allows for better control over composition (though it still works well with realistic models). Regrettably, it also handles clothing details less accurately—structural errors in garments may occur even at full strength (especially when used alongside other LoRAs). For this reason, I’ve kept the original version available for download, so you can choose the one that best suits your needs.

Files with SHA256: 3bee26e219358869dd44e5cc14754d6e876eff36490057a9e06e67976cccff4a