Type | |
Stats | 1,606 381 |
Reviews | (402) |
Published | Jul 7, 2023 |
Base Model | |
Hash | AutoV2 58024D4F41 |
这是做简笔涂鸦风格的lora,底模用的是seekyou_alpha1,
尺寸是512x768,描述词简单描述就好了,例如:狗,衬衫,红色鞋类,黑色短裤,短裤,1个男孩,黄色衬衫,简单背景,包,皮带,短袖,鞋,行走,袜子,简单的背景,短发,张嘴,棕色头发,背包,阴影,黑眼睛,腮红;
除了人物,也可以用在场景上,可以试试其他大模型用(偏二次元风格的底模比较适合);
权重看着调就好;如果勾选高分辨率2倍可能会出现结构错位的毛病,我一般高分辨率修复弄到1.3倍;迭代步数是20;
The following is a machine translation:
This is for a lora in the style of a sketchy doodle, the base die used is seekyou_alpha1.
the size is 512x768, the descriptors are simply described, for example: dog, shirt, red footwear, black shorts, shorts, 1 boy, yellow shirt, simple background, bag, belt, short sleeves, shoes, walking, socks, simple background, short hair, open mouth, brown hair, backpack, shadows, black eyes, blush;
can also be used for scenes in addition to characters, try other large models with (base models in the secondary style are more suitable);
weights look at the adjustment is good; if the checkbox high resolution 2x may be the fault of structural misalignment, I generally high resolution repair get to 1.3x; iteration step is 20;