SPICYBABESV3基本用掉了所有目前常用模型搭配,几近完美,具体情况看ABOUT THIS VERSION.
20230627------------------------------------------------------
第二版手的位置调整了一下,但是出图和测试效果不大一样,需要embedding负面词汇解决一下(例如站内搜bad handv5-neg或者bhands-neg)。模型比率调整到最优,画面清晰而且智能。
The position of the hand in the second version has been adjusted,But the results of drawing and testing are not the same, Need to embed negative vocabulary to solve it (such as searching for bad handv5 neg or bhands neg on the website),and the model ratio has been adjusted to the optimal level. The photo is clear and AI intelligent.
20230602split line
2.8D,much more realistic,merged by;
chilloutmix:0.5
babes:0.5
dreamshaper:0.5
cheesedaddy:0.3
aim at female much more colorful or clear face.
很多朋友来问关于生成图片正向词汇问题,答案是只是用于测试模型成图方向,涉及到模型扩散问题,具体也不是很清楚,可以用SD系统中图片反推/Tagger来找关键词,不一定准确有用。
Many friends come to ask about generating positive vocabulary for images(Short or missing text). The answer is that it is only used to test the direction of model mapping, and involves model diffusion issues. The specific information is not very clear. You can use image reverse inference/Tagger in the SD system to find keywords, which may not be accurate.