纯tag双人文生图并可以互动-Pure tag double humanities student map and can interact

这个工作流有什么用？不依赖lora也能同时绘制出两个大模型不认识的角色，在保持特征不会出现污染的情况下还能让两人进行互动，且操作非常简单，你只需要添加角色特征和互动动作即可，不需要什么复杂的操作。

不使用二次修图和CN，只使用文生图也能完成75%以上的互动，加上修图功能可以解决95%的问题。

不是只能生成第一张图片那种左右分开的，只是其他图像被删了，目前来看大部分姿势它都能够胜任，比如：磨豆腐、公主抱、扣扣等

这是很简单的工作流，使用简单的遮罩和一个第三方节点来达成目的，遮罩区域根据生图尺寸自动调节至图像的一半，不需要手动绘制遮罩（如果想，你可以更改它），有横向和竖向两种默认遮罩，可以按照构图需求来调用，使用时只需填写人物tag和动作tag即可，非常方便。这只是一个雏形，群友要的双人的，它的生成目前来看很稳定。你可以自由的在这个工作流的基础上进行更改和添加。我的目标是同时出五个角色，保持特征不被污染的情况还能指定其中两名角色为动作A，另外三名角色为动作B，目前还没有什么稳定的方案。

因为展示的图片涉及到了未成年，所以被删掉了很多，就目前以我的测试来看，它很稳定且简单。默认遮罩是对半分割的，所以生成左右构图的图片效果最好，同时我也做了其他方案可供选择使用，上下分割构图、调整比例来达到如“3：1”这样的构图、还有手绘遮罩。而且即使你要生成的构图是很复杂的，遮罩画起来很麻烦，那也不需要担心，我新加入了修图的节点，这是一个很好的解决方案。

注意看版本信息！！

Pay attention to the version information!!

What is the use of this workflow? Without relying on lora, it can also draw two characters unknown to the large model at the same time, and allow two people to interact while maintaining the characteristics without pollution, and the operation is very simple, you only need to add character characteristics and interactive actions, and no complex operations are required.

Without the use of secondary retouching and CN, Vincennes diagrams can complete more than 75% of the interaction, and the retouching function can solve 95% of the problems.

Not only can generate the first picture of the left and right separation, but other images have been deleted, at present, most of the poses it can be competent, such as ：grinding tofu, princess hug, hand sex, etc.

This is a very simple workflow, using a simple mask and a third party node to achieve the purpose, the mask area is automatically adjusted according to the size of the image to half of the image, no need to manually draw the mask (you can change it if you want), there are two default horizontal and vertical masks, can be called according to the composition needs, just fill in the character tag and action tag when using. It's very convenient. This is just a prototype, the group of friends want to double, its generation is stable at present. You are free to make changes and additions based on this workflow. My goal is to have five characters at the same time, with two of them designated as Action A and the other three as Action B. There is no stable solution yet.

Many of the images were deleted because they involved minors, but so far in my tests, it's stable and simple. The default mask is split in half, so it works best to create a left and right composition, but I also made other options to use, such as upper and lower composition, adjusting the ratio to achieve a "3:1" composition, and hand-painted masks. And even if the composition you want to generate is very complicated and the mask is very troublesome to draw, there is no need to worry, I have added a new retouching node, which is a good solution.