推荐分辨率:512*768
推荐采样方法:Euler a ,DPM++ 2M Karras ,DPM++ 2S a Karras,
如果开启高清修复,出图效果会更好~
高清修复推荐设置:
放大算法:R-ESRGAN 4x+ Anime6B 放大倍数1.5-2倍
重绘幅度:0.3-0.6
推荐大模型::sdhk_v30.preview,meinamix_meinaV11.preview,kidsmix_v10.preview,CounterfeitV30_v30.preview,darkSushiMixMix_225D.preview ,二次元模型基本都可以。
Lora触发词:niji,1boy,
Lora推荐权重:0.6-0.8
负面提示词:(EasyNegative:1.2), worst quality, low quality,
正向提示词可以写一些: horns, jewelry, heart, hood, sticker, earrings, cat, necklace, piercing, ear piercing, multicolored hair, jacket, short hair, pointy ears 或者一些颜色主图提示词,说不定是你喜欢的风格!!
我的b站名字:带你吃火锅儿 抖音名字:带你吃火锅
Recommended resolution: 512 x 768
Recommended sampling method: Euler a, DPM++ 2M Karras, DPM++ 2S a Karras,
If you turn on HD repair, the picture effect will be better ~
Recommended Settings for HD repair:
Amplification algorithm: R-ESRGAN 4x+ Anime6B amplification factor of 1.5-2 times
Redraw range: 0.3-0.6
Large models are recommended: sdhk_v30.preview, meinamix_meinaV11.preview, kidsmix_v10.preview, CounterfeitV30_v30.preview, Darksushimixmix225d.preview,
Anime models are fine.
Lora trigger words: niji,1boy,
Lora recommended weight: 0.6-0.8
Negative warning words: (EasyNegative:1.2), worst quality, low quality,
Positive prompts can include: horns, jewelry, heart, hood, sticker, earrings, cat, necklace, piercing, ear piercing, multicolored hair, jacket, short hair, pointy ears or some color main image cue words, maybe you like the style!