Karyln style

尽量复刻了karyln画师的风格.

用illustrious base模型训练,danbooru上能找到的大概140多张图,

挑选的最合适的模型共练了3400步左右.

实测在wai130和zukicute都有很好的效果.

I tried my best to replicate the style of the artist Karyln.

It was trained using the illustrious base model, with approximately 140 images found on Danbooru.

The most suitable model selected was trained for about 3400 steps.

It has been tested to have very good results on both wai130 and zukicute.

使用方法

触发词,Trigger prompts:

karyln_style,karyln style,flat color,sharp clean lineart,

负面词,negative prompts:

3d,best shadow,cinematic lighting,

目前没尝试出有什么其他更好的.(还是有高人指点了下才想起来需要符合风格的提示词辅助才行)

---

左边是生成图右边是原图

generated image and the original image.

此图可以大致的对比.非常巧合且意外地生成了一个动作几乎一致的图片.

This image can be roughly compared. A picture with almost the same movement was generated quite coincidentally and unexpectedly.

---

还是有些地方线条还原不够明显,不过目前足够了.

不使用完整触发词将会导致没有足够的特点.

There are still some places where the line restoration is not obvious enough, but it's sufficient for now.

Not using the complete trigger words will result in insufficient characteristics.