======================================================

Trianed with : NoobAI XL-VPred 1.0

Preview images were generated with : WAI-NSFW-illustrious-SDXL v14.0

I haven't tested the effectiveness using other checkpoints, so I can't guarantee the quality of the outputs generated by other models.

我没有使用其他大模型测试效果，所以我无法保证其他模型下的出图质量

Recommended weight 建议权重 : 0.8

Suggested :

• Sampler : Eular

• Steps : 30

• CFG : 7

Trigger words 触发词 : tinmanak

• Outfit 衣服 : necktie, kakhi fedora, kakhi hat, black leather shoes, belt, (kakhi trench coat), black pants, suspenders, white collared shirt, open coat,

• Common 通用：masked, flat face, smooth face, (without nose), (without nostril), half-closed eyes, black sclera, red pupil, black mustache, black sclera, bald, red eyes, silver colored robot, male android robot, robotic hands,

• Additional 附加：smoking pipe, paper bag,

• Negative Prompt 负面提示词：skull, nasal dorsum, nose, nostril,

======================================================

Some hints:

This lora model is trained based on 1 official pic, so there's might be a lot problem, for example: missing hat, looking weird and so on, so I spent more time on the prompt, if there's missing any objects or features, try to give more weight on the missing part's prompt like 1.3-1.5, e.g (kakhi hat:1.5).

这个模型是基于唯一一张官方立绘训练的，（没用精二立绘是因为背景太杂影响效果但是保留了精二立绘的头部特写加强特征），所以可能会有一些问题出现，例如缺少帽子或者看着很诡异，为了解决这些问题我花了更多的心思和时间在关键词上面，按照我给出的关键词生成，如果缺少某个特征或者物件，尽可能给予这个缺少的关键词更多权重，大概在1.3-1.5即可（尤其是使用Civit生成的用户，但是我还是不建议使用Civit直接生成 很多模型同样的关键词使用Civit出来的效果往往差很多的），举个例子：(kakhi hat:1.5)。

This lora model is trained with V-Pred model, even though it still can work with Illustrious models, but V-Pred model may not working very good with NoobAI EPS models such as HomoSimile XL and so on, so I can't guarantee the quality of the outputs generated by EPS models.

这个模型是使用NoobAI V-Pred模型训练的，是可以和IL的大模型一起用，但是可能不太适合与NoobAI EPS大模型使用比如HomoSimile XL，所以如果你使用EPS大模型去生图的话质量确实没法保证（摊手）