让高跟和紧身衣一体，不要让单独出来的高跟鞋毁了紧身衣的整体和谐性。
Make high heels Integrate with bodysuit and no more detached high heel shoes disrupt the overall harmony and integrity of the bodysuit.
穿紧身衣的女孩却光着脚，而穿的鞋居然跟美丽光滑一体的衣物分离了，真不舒服，更可怕的是被乳胶盖着的脚又外穿了一层高跟凉鞋真是我操了！我受不了这种暴行，所以我搞了这个lora来纺织有什么鞋或者靴子之类的出现在光滑的紧身衣上。
It is very uncomfortable to see the bodysuit girl has barefoot and wearing high heels detached from the beautiful, smooth and integrated suit, what even more irritating is the foot covered by latex and the fucking high heel sandals on it!!! I can't stand this outrageous situation, so I made this lora to prevent something in the shape of shoes or boots emerge on the smooth bodysuit.
Smooth bodysuit without any footwear:
Positive: heeled leggings,heelsuit,seamless,
Negative:
Smooth thigh boots:
Positive: heeled leggings,high heel boots,thigh boots,seamless,
Negative:
Description
FAQ
Comments (4)
Works fine for me.
Thx for sharing.
I will post some pictures to the gallery
Just a small prompting tip — that seamless heel style is called heeled leggings. In Krea2 you prompt it as: black latex leggings‑heeled‑seamless. You can support the lora using this in addition to suit.
I ran a few tests building token blocks, and for me the best results came from putting the color first, then attaching the other attributes with simple hyphens to the main character token. You lose a bit of semantic elegance, but Krea2 isn’t Flux — if you try to “explain” things to the model, you increase the chance of confusing it.
I am grateful to you for the valuable comments. I will tag it as leggings‑heeled‑seamless on next lora version.
@mengxinyimei If you want to do experimental training — which I’m absolutely a fan of — try something like:
“yourcolor high‑neck yourmaterial leotard‑long‑sleeves, leggings‑heeled‑seamless”
You can also hook the trigger right at the start, e.g.:
“S33ml33ss style”
Why this can work better than embedding the trigger inside the token block with the heel?:
The model tries to bind the trigger into its internal tag‑mesh. If the trigger sits too close to clothing tokens like leotard or leggings, it gets semantically fused with them. That makes the trigger behave like a clothing modifier instead of a style token.
By keeping the structure more open, you let the checkpoint handle about 50% of the clothing effect on its own. The trigger stays independent, and the LoRA doesn’t have to fight against checkpoint‑inbound clothing weights. That’s why you can often run the such a LoRA below 0.5 strength — the trigger isn’t overloaded. This will give you a wider range of creativity, cause your lora will not be a dominat player. Hope this helped you.
@gl4mdiva That's professional advice. Thank you very much. I'll use natural word in natural caption sentences, and put trigger word before the paragraph.