Holy fuck, this part is going to be hard to do with 3 different characters
>>45007589There is one that creates a heatmap of what each word affects and how much. What you have to understand however, is that AI doesn't really copy paste things from a database of hands and legs and faces. It is just performing a very difficult Rorschasch Ink Blot test multiple times until it gets from the equivalent of TV static to an Image looking like your prompt by sorting the static(latent image) into something that resembles your prompt more closely, each step.
There is also another tool that can decide if an image is visually appealing or not, ranking its aesthetic. And the most powerful one I've seen yesterday is a mass auto-mask creator that can mask images based on what is in them and avoiding specific things. Let's say you have an animation set with a girl singing a trumpet, and you only want to change how the girl looks like, you can tell the auto-masker to mask the girl but not mask the trumpet, without even opening up photoshop or touching a drawing tool.
