This workflow includes three guiders. You can test the effects of different guiders on the generated images. I recommend you to use DualCFGGuider. This guider allows you to fill in negative prompt words without affecting the image quality. You can mix styles by modifying CFG.

DualCFGGuider rules:

When cfg_cond2_negative is less than 1.0, the positive prompts's effect is weakened.

When cfg_cond2_negative is 1.0, negative prompts has no effect.

As the cfg_cond2_negative coefficient increases, the negative prompts's effect increases.

Negative prompts are more practical: determine age range, reduce freckles, change style, etc. In short, negative prompts make the image more controllable.

For the Schnell model, when the CFG of the negative prompts is greater than 1, strange images may appear.

What's more, if you want a realistic style, add “A photo shot by camera” at the beginning of positive prompts will be more effective.

//---------------------中文说明--------------------------

本工作流包含了三个引导器，你可以测试不同的引导器对生成的图片的影响。我推荐你使用 DualCFGGuider，这个引导器允许你填写负面提示词，而不会影响图片质量。你可以通过修改 CFG 来混合风格。

DualCFGGuider rules 规律：

cfg_cond2_negative 小于1.0，positive 提示词作用被削弱。

cfg_cond2_negative 为1.0时，negative 提示词不起作用。

随着 cfg_cond2_negative 系数增加，negative 作用增加。

负面提示词比较实用的用法：确定年龄范围、减少雀斑、改变风格等。

用于 schnell 模型，负面提示词的CFG大于1时，可能会出现奇怪的画面。

另外，如果想要写实的风格，在正面提示的开头加上“A photo shot by camera”会更有效。