自己训练的第一个Lora,使用的是朱尼酱的赛博炼丹器制作,对于模型有任何建议都可以评论,欢迎返图。
因为训练集的问题,有时生成的图片会有汗水,可以在负面提示词里添加汗水的英文sweat
模型的权重推荐0.6-0.8,也可以搭配其他Lora使用,效果请自行摸索。
参考图比较模糊是因为是本地生成的756*512尺寸的原图,可以后期放大处理
The first Lora I trained was made using 朱尼酱's cyber alchemist, and you can comment on any suggestions for the model, welcome to return to the picture.
Because of the problem of the training set, sometimes the generated pictures will have sweat, and you can add sweat to the negative prompt words
The weight of the model is recommended to be 0.6-0.8, or it can be used with other Lora, please explore the effect by yourself.
The reference image is blurry because it is a locally generated 756*512 size original image, which can be enlarged later