 So you don't know how to keep your characters consistent, not a problem. This is probably one of the most common challenges that AI artists are going to face in the future. And there are a few different ways to manage this, but the first one is to take advantage of the default characters that are already within Dolly. One of my first projects in Dolly was making a movie poster for a friend of mine who works at Starbucks. And just about every time I typed in Starbucks movie poster, I basically always got a guy that looked like this. Now this is a consistent character that pops up when you type in Starbucks barista movie poster, which means that he generally represents the average characteristics of most Starbucks barista pictures. And that profile is dark hair, dark eyes, fair skin, and wearing an apron. Which means if we add the word female in front of all this, we should get a girl with dark hair, dark eyes, fair skin, and wearing an apron. And yes, that is essentially what we get. Awesome. So we can get consistent characters, but how does it work? Well the reason is pretty simple. If your description is too short like girl, you're basically going to get a random image each time because you're not giving it any specifications. The reason why people struggle to get the same looking character is because they try and brute force the prompt with way too many descriptions. They'll say something like, I want a young girl with pale skin, soft features, dark hair, dark eyes, and wearing a blue dress. And you have to understand that every word in the prompt is a cluster of related words. And often the more words you have, the more relationships there are that you might not anticipate. So in order to avoid that, we can hijack the system by front loading as much information as possible with as few descriptors as possible. So instead of manually describing pale skin, soft face, dark hair, dark eyes, wearing a blue dress, instead all we really have to do is find a word that by default contains all of these relationships front loaded in. For example, we could replace all of this with Japanese. Because Japanese girls 99% of the time will have pale skin, soft features, dark hair and dark eyes. And the AI automatically knows all this without you needing to describe all of it. So now if we search, you can see that the images are much more consistent than they were before. Because even though the description is shorter, the shorter description contains much more accurate information related to our target and has fewer word clusters, so there's fewer unexpected relationships. The same way if I needed to create a character with blonde hair and blue eyes, instead of describing all that, I can take a shortcut and describe all of that info. If I just say something like German girl or Russian girl or Swedish girl. And the final result will automatically inherit those common characteristics that are most associated with that word cluster. So it's more likely to be consistent than if you were to type everything manually. Every time me and my team finds the same character, we call it a Dolly OC. And Dolly actually has a lot of OCs that you can trigger with the right words. For example, if you type in T Rex, you're going to get all kinds of different T Rexes. But as soon as you say T Rex riding a tank, the design for the T Rex is always going to be consistent and green. If you type in steam punk pirate king, you're always going to get this guy. Gamer grandma pretty much always looks like this. A young girl in a blue dress playing with a German shepherd will usually look like this. Oh yeah, angels are a great character. They're pretty consistent. And speaking of which, Jesus is also a very consistent character that you can use. And to take it a step further, girl Jesus is an excellent reoccurring consistent character in Dolly. She's probably one of my favorite reliable archetypes for generating anime brunettes. So I think you get the point. The average of the word cluster is almost always the same. So when you're describing a character, front load your prompt with existing concepts or characters that describe what you want in a smaller amount of words. So hope that helps Santa as always. Hope you have a fantastic day and I'll see you around.