这个模型是融合模型,感谢hans,浅笑大人,青秋,天时与人和(排名不分前后)大佬的支持!
上色风格多数情况下为半厚涂,减少质量词和负面TAG会偏向平涂,整体上色风格较为鲜艳,欢迎尝试~
建议参数:
CLIPskip:1~5(推荐值为2,如发现画风崩坏再尝试更改该参数,不建议超过5)
提示词引导系数CFGScale:7~13
图片大小:建议宽高至少有一个大于等于768,且宽高比在0.5~2之间,推荐512x768、768x768、960x640、768x1024、1024x1024
高分辨率修复:建议使用4x-UltraSharp、R-ESRGAN4x+Anime6B,放大倍数2~4,重绘幅度0.2~0.5
使用简单反向可以改变平涂/半厚涂
建议反向提示词Negative prompt:
bad anatomy,cartoon,anime,sketches,extra hands,bad legs,wrong legs,bad shoes,bad arms,extra legs,bad feet,bad shoes,messy hair,extra arms,extra limbs,missing limbs,short upper-body,skinny,cloak,dress hem,extra arm,(wrong clothing),(extra hands),messy clothing,bad arms,wrong clothing,blurry eyes,bad shoes,bad legs,bad shadow,(bad anatomy),wrong body_structure,(bad hands),(worst quality, low quality),(deformed, distorted, disfigured),(bad eyes, wrong lips, weird mouth, bad teeth, mutated hands and fingers:1.2),bad-hands-5,badhandv4,negative_hand-neg,easynegative,FastNegativeV2,verybadimagenegative_v1.3,ng_deepnegative_v1_75t,