Type | |
Stats | 438 1,073 |
Reviews | (90) |
Published | Apr 15, 2024 |
Base Model | |
Training | Steps: 20 |
Trigger Words | kk-boy |
Hash | AutoV2 F41FDB41A8 |
如果你喜欢这个模型就点个赞吧，您的支持就是我最大的动力~
因为国内关于男性的模型比较少，很难找到让我比较满意的男性Lora模型，所以就自己练了一个...
主要争对男性人物的绘制，但在画风上也会有一定作用。
推荐配合大模型：AWPainting V1.3（其他大模型也可以）
触发词：kk-boy（其实不加也行，因为没有进行正则化，所以像1boy，solo这样的词都被污染了）
推荐lora权重：0.6~0.8，可以根据需要自行调节lora权重。
因为训练尺寸只有512，所以生成一些小饰品或者手脚啥的会糊，推荐开启插件高清修复，但更建议后期PS处理
迭代步数Sampling Steps 25-30
采样器推荐设置 Euler a
高清修复推荐设置 R-ESRGAN 4x+ Anime68/ 4x-UltraSharp, 重绘幅度0.3-0.4
经过测试能比较好的与其他Lora模型很好的结合，欢迎大家继续尝试！
有好看的图片欢迎大家返图啦！
If you like this model, please click "like", your support is my biggest motivation
Because there are few models about men in China, it is difficult to find a male Lora model that makes me more satisfied, so I practiced one by myself.
The main contention is the drawing of male figures, but it will also have a certain role in the painting style.
Recommended for large models: AWPainting V1.3 (other large models are also available)
Trigger word: kk-boy (In fact, it is OK not to add, because it is not regularized, so words like 1boy, solo are polluted)
Recommended lora weight: 0.6~0.8, you can adjust the lora weight according to your needs.
Because the training size is only 512, so the generation of some small ornaments or hands and feet will paste, it is recommended to open the plug-in HD repair, but it is also recommended that the post-PS processing
Sampling Steps 25-30
Euler a is recommended as the sampler
The recommended setting for HD repair is R-ESRGAN 4x+ Anime68/4x-UltraSharp, with a redraw range of 0.3-0.4
After testing, it can be well combined with other Lora models, welcome to continue to try!
There are beautiful pictures welcome everyone back to the map!