 So once you've picked your style model, the next thing we need to do is caption your data set. Now there are two ways to do this, the automatic way, which is easy and works pretty well, and the manual way, which does give you some better results, but I always think it's good to know both. So today I'll just show you how to use the auto caption tool in Koya. Now something you'll need to figure out is the keyword that you would like to use to identify your character in stable diffusion. The keyword is what you use to summon your character whenever you want stable diffusion to generate it. Now this word can be anything and it's actually best if it's not really a word more like an acronym or a combination of numbers and letters, whatever you come up with, make sure your keyword is not already taken by something else. So it's good practice to type in your keyword ahead of time and see what gets generated. If the results are random, or they give you a variety of different things, then that is usually a pretty good sign. If you type in something and you always get the exact same character, it probably means that if you train with that keyword, you run the risk of your character mixing too much influence from what's already attached to it. For example, if you use a keyword like cow boy, but you're training a female character, you probably run the risk of generating both male and female versions of your character because your keyword is very closely related to the word cowboy and boy. So you're probably better off changing the keyword to be something more like cow girl or C girl or C G R L something something that isn't the opposite of what you are hoping to train. But the basic rule is even if you see other characters come up, as long as they don't always generate the same clothing, face, hair, eye color and background all the time, then you're probably fine. For my character, I'm just using the keyword AHRA, which is just an acronym. And by default, this keyword generates these kind of results. None of these characters look like they are over trained. They have different faces, different hair and eye colors, and they're all attractive females with a somewhat similar build to the kind of character I'm trying to train. Regardless, once you have your keyword picked, go to your Koya training file and add a new folder for your character, open it, then create three new folders named image, output and log inside the image folder, create a new folder named 20 underscore and the keyword for your character, and then drag all of your training images into this folder. Now if you're wondering what this number is, what I'm about to say is a gross oversimplification. But basically, this number determines how many times the training data runs through all these images. Essentially, the training will go through 20 sets or epochs of this folder. Usually 20 is a pretty safe number if your training data set is fairly small. So assuming that you have 20 to 30 images like I do, then the number 20 is a pretty safe number of sets. Once you've done that, launch the Koya GUI that we installed a few videos ago, go to the utilities tab, find the folder with all your training images and create your auto caption prefix here. And this should just be a general description of what your images contain in my data set. Basically, we just have the keyword. And what we see is basically a woman and a green background. And we're just going to stop there because the simpler you keep it, the more flexible the training can be. Now, sometimes if there are important details about your character that you might want to change, for example, my character is wearing a hat. And if we wanted to make sure stable diffusion noticed that then we could add a hat to this caption. And this would help us later on ask it to do things like remove the hat or hold the hat. But I personally know my character will always be wearing a hat. So I don't think that's a detail I really want stable diffusion to hyper focus on. So we're just going to keep the description like this, make sure you check overwrite existing captions, click here, you're done. Congratulations. Now you know how to use the auto caption for your training data. Hope that helps. And as always, I'll be of a fantastic day and I'll see you around.