它在512x768训练,It trains at 512x768
v1.1===============================================
示例图的放大过的图片连续跑了一些不同场景的，可以稳定的大概率的得到这种姿势。
我改进了原来的版本,删减了一些关键词并增加了训练度,使用更少的tag来触发这个姿势,accidental exposure panties, seat使用这两个来触发效果，如果没有闭眼睛，请给accidental exposure panties加权重或者加closed eyes。
我展示了xyz图,在写实，2.5d,和动画模型的效果。也许我的触发词设置的不好，动画模型完美理解了这个姿势。
它可以换一些地方，自行尝试把，但我不能让她去月球😂
The enlarged image of the sample graph runs consecutively through several different scenes, which can be achieved with a high probability of stability.
I improved the original version by deleting some key words and increasing the training degree, using fewer tags to trigger the pose,'accidental exposure panties and seat' to trigger the effect, if the eyes were not closed, Please weight accidental exposure panties or add closed eyes.
I showed the effects of the xyz diagram, in realistic, 2.5d, and animated models. Maybe my trigger word was set up badly, but the animation model understood the pose perfectly.
She can go somewhere and try, but I can't let her go to the moon😂
v1================================================
加入这些：accidental exposure panties, short skirt,train,sleeping,closed eyes,
剩下的看你发挥
Join these： accidental exposure panties, short skirt, train, sleeping, closed eyes,
The rest is up to you
Description
FAQ
Comments (11)
Train only? More locations would be nice
Thanks for your proposal, I improved it in the new version
@wwwadadad564 Awesome, thanks!
Doubtful lora
Thanks for your proposal, I improved it in the new version
Works perfectly, well done
Does it work in real, like 3d
thank you for cool concept!
from your xyz graph, it looks like the LORA makes realistic Chilloutmix turn into anime?
is there anyway to keep it realistic? I think version 1.0 is slightly more realistic than 1.1? do you have a version thats fully realistic?
google translate:
谢谢你的酷概念！
从你的 xyz 图中，看起来 LORA 使逼真的 Chilloutmix 变成了动漫？ 无论如何要保持现实？ 我觉得1.0版比1.1版稍微现实一点？ 你有一个完全现实的版本吗？
xyz只是一个参考，我在1.1中所有的样品图片都是写实模型。如果还不够写实你可以试试1.0样品图片引用的一个lora，它的名字叫'胶片女孩'，它对于真实图片的效果有所帮助
1.1和1.0没有什么大的改变,只是我加重了学习的强度，能够更容易的产出这个姿势。
xyz is just a reference, all my sample pictures in 1.1 are realistic models. If that's not realistic enough you can try one of the lora references in the 1.0 sample images. It's called 'Film Girl' and it helps with the real image
There is no big change between 1.1 and 1.0, but I have increased the intensity of learning and can produce this posture more easily.
I like this style very much. I wonder if you can only extract this style for one lora?(hope)(hope)
It's only accidental cuz the eyes are closed, otherwise its just another upskirt lora