Recommend including "monochrome" in prompt when using it. (But if you don't, results may be interesting.)

Trained with over 100 pictures, all earlier than 1922, mostly created by Wierix family and Gustav Doré, along with a few others. I chose mainly portraits and scenery for training. (So there weren't many angels or demons, heaven&hell thing.)

Base model is SD1.5

Besides SD1.5, I also tried using this lora with models that are more 'anime', like Anything 4.0. It works a little, but hardly any style change to the face. I dunno, maybe that's okay?

I am not very satisfied with this lora, if you know how I can improve it, I'd like to hear.

推荐搭配“monochrome”使用，更稳定一些。不加的话也可能会产生奇妙的色彩效果。

使用的训练图片均早于1922年，大部分是Wierix家族和古斯塔夫·多雷的作品， 并且主要是人像和自然景物（基本上没有多雷的神曲和失乐园的插图）。

训练基于SD1.5，并且跟SD1.5搭配的效果比较复古；而如果在偏二次元的模型上使用……这个lora似乎难以改变二次元美少女的脸，我说不准这是好事还是坏事（

我对这个模型的效果还不太满意，正在寻找改进的方法。