这篇帖子是《胜利女神：妮姬》角色cosplay服装FLUX.1 D的LORA集合(包含官设和同人)。
This post is a collection of "GODDESS OF VICTORY: NIKKE" character cosplay costumes FLUX.1 D LORA (including official and doujin).
触发词/Trigger word (示例图片触发词/Sample images trigger words):
请参考各个版本的说明
Please refer to the description of each version
LORA采用黑森林原版flux作为底膜进行训练。使用时LORA强度可以直接设置为‘1.00’
LORA uses the original Black Forest Labs FLUX as the base model for training. When using it, the LORA Strength can be directly set to ‘1.00’
这个模型主要聚焦在服饰，脸部不在预期训练范围内。你可以搭配其他脸部lora或者使用facedetailer来进行脸部重绘。效果图我使用了FaceDetailer进行脸部重绘。
效果图我的出图流程是:
FLUX官方演示工作流直出 1152*1536 -> SD1.5脸部重绘（个人比较喜欢sd1.5的脸，您可以根据您喜好重绘）
This model mainly focuses on clothing, and the face is not in the intended training range. You can use face Lora or use facedetailer to repaint the face. I used FaceDetailer to repaint the sample images.
My workflow for sample image is:
FLUX official demo workflow generate 1152*1536 iamges directly -> SD1.5 FaceDetailer (personally I prefer sd1.5 face, you can repaint it according to your preference)
诚挚邀请您尝试此LORA，作者很希望能得到您的反馈和建议。如果您能分享您生成的美图就更好了。感谢~
Sincerely invite you to try this Lora. I am eager for your feedback and comment. It would be great if you can share the beautiful images generated by this lora. Thanks~
Description
触发词/Trigger words:
A photo of a beutiful tall slim woman in nikke-blanc cosplay costume. She is wearing long sleeves cutout japanese clothes and kimono with dragon print and sash and black pantyhose and black high heels. she has white hair with hair bun and hair ribbon and hair ornament.
示例图片触发词/Trigger words of sample images:
A photo of a beutiful tall slim woman in nikke-blanc cosplay costume. She is wearing long sleeves cutout japanese clothes and kimono with dragon print and sash and black pantyhose and black high heels. she has white hair with hair bun and hair ribbon and hair ornament. She is standing on the edge of the roof with neon and led lights, behind is a futuristic cyberpunk skyline with skyscrapers. The photo is a full shot with her entire body visible and full body and with detail background.
FAQ
Comments (9)
请问您训练这个模型的数据集一般是准备多少张呀？还有训练步数要多少步？
一般 20 张以上吧，步数在 2000 步左右
用FaceDetailer修复后，越修越丑，我实在是绷不住了。
😂 不应该呀，FaceDetailer挺强的
@cyberAngel_ 挺难调的，不同模型调起来效果不仅差别大，还很不稳定，就好像脸和五官不配套那种感觉。
我使用您的lora时，衣服上的文字出现小部分，细节也不能保持大体一致，想请问一下样图是在细节上有后期重绘处理还是我的某些设置没有做得很好，或者是我的底模并没有使用黑森林工作室的？
文字出现小部分错误，我想达到您样图的效果来保持一致性，求教！
这个问题我已经解决，采样器的问题，大佬可以的话考虑一下八神虫子（原神），雷电将军（原神），山里凌华（原神），申鹤（原神），宵宫（原神）纳西坦（原神）妮露（原神）这些出得早的，约会大作战，西施游龙清影，西施原皮，妲己青丘九尾，公孙离祈雪灵祝，影；动漫的斩赤红之瞳那些老牌动漫，圣斗士星矢，假面骑士都请参考一下！！国产的一代铠甲勇士也可以！
@suqingyou5412 好滴，看到好的素材都会练滴，谢谢建议~
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.