V2：

抽空把管理员更新了一下

由于素材比上次多了很多很多（上次只有一张素材

所以稳定性极大幅度提升

除了因为素材多了导致小日子的画风被冲淡，黑丝质感可能比不上V1

就没有别的缺点了

当然，衣服特征还是没法完美还原

与其指望这个LoRA能把她身上零零碎碎的配饰都画出来不如去试试NAI3

请使用新版的tag，下面那些都是V1的

顺带一提，最近一段时间应该不会再继续训练LoRA了

不如说以后sd1.5的LoRA都没有意义了也说不准

Took some time to update the Endministrator.

Due to a lot more source materials compared to last time (only had one piece last time),

so the stability has greatly improved.

Apart from the fact that the increase in materials may dilute the art style of the character, and the texture of the black stockings may not be as good as V1,

there are no other drawbacks.

Of course, the clothing features still cannot be perfectly reproduced.

Instead of hoping that LoRA can draw all the scattered accessories on her, it's better to try NAI3.

Please use the new version of tags; the ones below are all for V1.

By the way, it seems that LoRA won't be trained for a while.

It's uncertain whether there will be any meaning for LoRA based on sd1.5 in the future.

V1：

又是因为没人发所以我就自己练了一个

虽然pixiv上貌似已经有人发过她的ai图了，但是并没有公开模型

所以现在就由我来给大家开源

全部的素材都由官方发的唯一一张立绘裁剪得到，所以别指望着得到的图片能有多精致

因为她的衣服上本身就有大量的细节，而目前可用的素材又屈指可数，所以想要画好细节会非常困难

等之后同人图多了之后会考虑更新，稳定性应该会得到不小的提升

目前来说，对于只有一张素材的LoRA来说，这个模型的表现已经算好的了

推荐权重为0.8左右

极度推荐使用LoRA Block Weight进行分层控制，推荐预设为1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0

关键词为Endministrator, open coat, hairclip, black_footwear, sweater_dress, black_pantyhose, black_coat, black jacket, long_sleeves, windbreaker, loafers, short_hair, grey_eyes, blunt_bangs,, black_hair,

请根据实际情况调整这些关键词的权重或者考虑是否删减一部分，这取决于你的需求

可以使用x hair ornament使得头饰更接近立绘中的款式，尽管效果仍然不太理想

毛衣裙配黑色裤袜实在是太戳我的性癖了

If you like this LoRA, could U plz Click the FOLLOW?

It's because no one else posted it, so I created one myself.

Although it seems that someone has already posted her AI image on Pixiv, the model has not been made public.

So now I will open source it for everyone.

All the materials were obtained by cropping the only official illustration, so don't expect the images to be very refined.

Because her clothes already have a lot of intricate details, and there are only a few available references, it will be very challenging to draw the details well.

I will consider updating it once there are more fan arts, which should improve its stability.

For now, the performance of this model is already considered good for LoRA, which only has one material.

I recommend a weight of around 0.8.

I highly recommend using the LoRA Block Weight for hierarchical control, with a recommended preset of 1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0.

The keywords are 'open coat, hairclip, black footwear, sweater dress, black pantyhose, black coat, black jacket, long sleeves, windbreaker, loafers, short hair, grey eyes, blunt bangs, black hair.'

Please adjust the weights of these keywords based on the actual situation or consider whether to delete some of them, depending on your needs.

You can use x hair ornament to make the headpiece closer to the style in the illustration, although the effect is still not ideal.

The combination of sweater dress and black pantyhose really hits my fetish.

描述:

除了因为素材添加了很多别的画师的作品，导致小日子的黑丝质感可能比不上V1

除此之外全方位碾压

但是衣服上的细节肯定还是没法完美生成的

与其指望这个LoRA能够把衣服上的细节画完整不如去期待NAI3

训练词语: Endministrator_(Arknights), short_hair, black_hair, grey_hair, white_sweater, sweater_dress, grey_shirt, turtleneck_sweater, black_jacket, open_jacket, short_shorts, black_pantyhose, black_footwear, hairclip,