 So you have no idea how prompts interact with the results. Not a problem. If you wanna learn to engineer really good prompts, it helps to know what the limitations of the machine are. So for example, let's see how many things we can throw at the machine until it overloads. So let's start with something simple like beautiful young fantasy princess. Okay, now let's add a blue dress. Okay, good so far. Now let's add a prop. So something like an umbrella. Okay, this is where the machine starts to show its cracks. Usually it gets it, but it's starting to put the umbrella in the wrong locations. Let's see what happens if we try and type in a yellow umbrella. Now it's getting really confused making her yellow, sometimes making the umbrella blue. Let's see if we can fix that with punctuation. Okay, commas don't really seem to affect it. Let's see if we change these commas to periods to see if that makes a difference. Doesn't really seem to make a difference. So the takeaway from this is you'll get your subject almost 100% of the time and you can add a prop. But as soon as you start describing the prop, you're getting into wishy washy territory. There's no guarantee that what you describe is gonna connect beyond one subject and an object. If we try to add something after this like a pink castle, then yeah, the machine gets all confused. The castle's the wrong color. The dress is the wrong color. Sometimes the prop is the wrong color. Just be prepared to get only two out of the three things you ask for. This is actually one of the few places where engines like Dolly and I suspect Google Party are going to have the advantage over stable diffusion. If we type a watercolor painting of a beautiful young fantasy princess holding a red flower in front of a castle during sunset, trending on art station, you can see it generally does a pretty good job. We got the princess, we got the red flower, we got the castle and we even got the sunset. So hopefully stable diffusion will improve on that pretty soon. So now that you know the limits of the machine, how should you organize your prompt? Well, the best format that I found is you start with your media. So if it's a portrait or a painting or a photograph, that stuff goes in the front. The next part is your subject. You can try to add an object, but just remember, no more than two things. Princess holding a flower, princess in front of a castle, but stick with two. And then after that comes all your descriptors. Usually these are separated with commas. Some of the best, most common descriptors are beautiful, delicate, ultra detailed, attractive, young, illustration, smooth, sharp. These are the kind of descriptors you want to use if you want to get something pretty. And then at the very, very end of your prompt should be the artist or the style that you're trying to emulate. So if you want it drawn by Claude Monet, that goes at the very end here. And you can also mix different artists together. So if you want to combine Claude Monet with Alfonso Mucha, you can do that. By far the most common combination that everyone likes is Art Germ, Greg Rutkowski, and Alfonso Mucha. You can never go wrong with these three. So just remember the format is type of media, subject object, descriptors, and the artist. That's your format. So if you're trying to generate an image of a princess in a blue dress holding a flower, I would probably word it like beautiful, cottage core, fantasy, young, blue Victorian princess holding a flower, full body shot, intricate, elegant, highly detailed, digital painting, trending on ArtStation, concept art, smooth, sharp, focus, illustration, by Art Germ and Greg Rutkowski and Alfonso Mucha. Max out the sample rate for the best quality and there we go. So that's how prompts work. Hope that helps and as always, hope you have a fantastic day and I'll see you around.