Ep. 18: Actionable ChatGPT Tips to Make Midjourney Even Better

0:00:00Intro

Play from 0:00:00 8:88:88 0:00:00 [00:00:00] And we're back. You like that? That was good. Midjourney Fast Hours Podcast. Drew Brucker, Rory Flynn. What's happening? Oh, nothing much, man. We are closely heading towards the holiday season. Don't know [00:00:15] how. I like looked up and I was like, damn. It's like two weeks away from Thanksgiving. How the hell did that happen? [00:00:20] So, it's literally flying by. Having, having like a subtle meltdown, cause, you know, Black Friday is always absolutely insane, and then I realize that it's, you know, gonna do this for the [00:00:30] first time with a, uh, with a baby, so we'll see how this goes. But, you know, man, how's, how's, uh, how's your life been going so far?

Play from 0:00:36 8:88:88 0:00:36 [00:00:36] Give us a little rundown. Hide your credit cards from your wife for Black Friday. Jesus. Yeah. I, uh, what's, what is going on, bro? I don't know. Just, just [00:00:45] busy. creating some new content with, uh, with Midjourney, trying to play with some new tools. Uh, you know, just everything, man, everything we've been doing. [00:00:52] Um, You know, really excited about what's coming and just all the changes that are happening, right? it's also challenging, Because [00:01:00] for us, I know we're talking to a lot of people that, maybe our clients or we're, we're pitching work to, you know, and to keep them up to date, right at different foundational levels can be tough to explain what's possible, what's not possible.

Play from 0:01:11 8:88:88 0:01:11 [00:01:10] And you also have to be up to date with that stuff. So, yeah, man, I mean, like, you know, [00:01:15] that, that's sort of a challenge, but, uh, in general love this time of the year. This is a. I think this is probably like my favorite time, you know, like football season, college basketball season, you can finally maybe wear a hoodie. [00:01:28] If Atlanta is not too fucking [00:01:30] crazy, you know, like it's 80 today. Like, can we just stay in the sixties for a couple of days? So I can actually, you know, not have to change middle of the day. So that'd be great. I don't know how you feel. I was at. I was in Charleston this week, quick trip, with the baby, first flight.

Play from 0:01:45 8:88:88 0:01:45 [00:01:44] [00:01:45] Yeah, how was that? She performed well. She performed well. I'll give her, you know, we were like giving her a solid like 83 out of 100 grade on that. You're going, you're going on the full scale system, not even a 1 through 10. Okay. Well, you know, we were, we were sitting there, we're like, all right, she's being really [00:02:00] good as the plane started to take off. [00:02:01] And I'm like, You know, everyone's like, feed her during takeoff, or so her ears don't pop. And I'm like, if we're feeding her, that means she's gonna poop herself. And, you know, damn well what happened as soon as we hit that tarmac. Start hearing the Pbbbt! [00:02:15] It's like, oh no, oh no! Here it is! So, you know, she's just like, smiling, like, sitting in her own, you know, self.

Play from 0:02:23 8:88:88 0:02:23 [00:02:23] And then on the way home, she was, we had a little bit of a meltdown on the tarmac. And I was like, oh god, everyone's gonna kill me and like, want to slit my throat. And then as [00:02:30] soon as the plane started lifting off, she was like, Khh! Out. It's amazing. I think. But. Yeah. It was like, 85 down there. [00:02:37] I was like, holy, like, I, we were walking around like, sweating and dead. I was like, I haven't felt It's humid! Dude, it's In a month. Right? Yeah. It's super humid still. Yeah. [00:02:45] Mm. It's wild, man. It's wild. What did you guys do out there? We were, uh, Yeah. We kind of just like, you know, I went to school in University of South Carolina, so I'm pretty well versed in Charleston and going, you know, been there a bunch of times seen, you know, ate, eat in a bunch of restaurants.

Play from 0:02:59 8:88:88 0:02:59 [00:02:59] So we kind of just, [00:03:00] uh, we're like, what are the, we're the low key spots that we can go and take the baby. They got some good food. They got some really good food. So good. We had this place one 67 raw for. Lunch one day, you know, we just started piling this this is like the difference between having a baby and just [00:03:15] going by yourself You know, my wife and I normally probably like get a bottle of wine have an awesome afternoon get hammered maybe like, you know, just walk around just be idiots throughout the city and You know baby wakes up when we're at lunch So it's like I'm eating oysters with one hand bouncing the baby with [00:03:30] the other You know, it would switch, like, pass to the wife, wife starts, you know, feeding the baby with the bottle, I'm like, hammering down a meal, she's like, switch, back over to me, so it's like, you know, just try to keep her in, you know, in relatively high spirits, so that we don't have to, like, ruin our lunch, [00:03:45] and, uh, That was kind of what it was like the whole time.

Play from 0:03:47 8:88:88 0:03:47 [00:03:46] But I hate, you know, you, you get master multitasking, right? I get it. I get it too. Do you even go out to restaurants? We don't switch. Uh, so you had one in one. There you go. Like my [00:04:00] strategy now is as soon as I order the food, I'm telling them to bring the checkout with the food. Right. So I can already have that ready to go. So as soon as it's done, we jet. And it's crazy because every time you ask that, like you catch. them off guard. Yeah. And I feel [00:04:15] like they think it's a knock on them for some reason.

Play from 0:04:17 8:88:88 0:04:17 [00:04:17] And I'm just like, nah, like you see what's going on here. Like, they don't get the chance to upsell or boost that tip a little bit. Yeah. Glass of wine, extra drink. That's what it is. There you go. Yeah. That's what I remember the days I was a bartender. You know, it was [00:04:30] always like a still sneak an extra little, Oh yeah. [00:04:32] How about dessert? It's really good. And then it's, but the, uh, the funny thing is with that, I saw someone, it was, uh, It was one of my, one of my bosses, Lisa, she came to visit us in New York and like she brought her, [00:04:45] uh, I think he was five at the time and like, as soon as we sat down, she had like, they were like, Hey, how are you doing?

Play from 0:04:51 8:88:88 0:04:51 [00:04:51] She's like, can I have pasta with butter and orange juice, this and that and bring that out just immediately. Then, you know, come back and take our drinks, like whatever. She was like, just get the food in their hands. [00:05:00] He'll eat it, he'll be happy, and like, I don't deal with it. It changes the whole experience, dude. [00:05:04] It's an entirely different thing. Well, at least you made it, tested it out, you know, know it's possible. Survived. The things that, uh, things that you don't think about [00:05:15] is like, Oh damn, like, the little walker and the bouncer provide so much entertainment. Even if it's for like 20 minutes at a time, like you need that 20 minutes to, I don't know, go to the bathroom or like, you know, get dressed or take a shower every [00:05:30] once in a while.

Play from 0:05:31 8:88:88 0:05:31 [00:05:30] So it's like, you know, you just constantly holding, constantly, you know, moving her around, standing her up, sitting her down, laying her down, playing with her. Hey, this is the real stuff, guys. She'll be back soon. You guys will hear her as normal, don't worry. She'll be [00:05:45] screaming in the background doing something. She's got a raspy voice right now. It's actually really funny from the plane. I think from the dry air. It's like, ehhhhhhh. Oh my god. That is hilarious. That's like a grandma. Holy shit.

Play from 0:05:57 8:88:88 0:05:57 [00:05:57] That's great. Oh man, we might have to, [00:06:00] yeah. This is what, this is what goes on in between Midjourney generations. So, Speaking of Midjourney generations, Let's talk about the topic for today, because I think this is a good one. Uh, we touched on, this a little bit back in, I don't [00:06:15] know, episode five, way back toward the beginning, but it was in combination with a bunch of other tools, right? [00:06:20] So today we're going to be talking about ChatGPT or your LLM of choice, right? It doesn't have to be ChatGPT, but, um, you know, how you [00:06:30] can maybe use that in tandem with Midjourney. And of course there are some obvious ways to do it. we'll probably touch on a few of those, but there's also some unconventional ways to do it.

0:06:15ChatGPT Basics

Play from 0:06:40 8:88:88 0:06:40 [00:06:39] And it's hard to really say how much people are using ChatGPT in combination with Midjourney, [00:06:45] because, you know, we don't, we don't really get to sit with too many people and just kind of watch their process, but, um, hopefully this unlocks something for you because there are so many different valuable ways. [00:06:56] To leverage it in addition to [00:07:00] creating jerk images, assessing images, uh, describing images, putting into the use cases, all that good stuff. So we're going to dive into that. anything you want to get us started with? I think it's really important that, you know, again, this isn't like, uh, here's, here's 50 prompts for Midjourney with chat [00:07:15] GPT.

Play from 0:07:15 8:88:88 0:07:15 [00:07:15] It's more so of like a mindset and a thought process of going about it. Because again, that's why these tools are awesome. That's why having ChatGPT, Claude Lama, whatever, they're legit because they're just problem solvers. So you can get creative with them. Also. It's not just [00:07:30] like, here, can you create, help me create prompts for my images? [00:07:32] Yes. We'll go through that probably a little bit here, but also like there's different ways to use it and I think it's, you know, again, how, how creative can you get with the. Infinite possibilities that we have with the tools and, you know, also just get better and learn.

Play from 0:07:45 8:88:88 0:07:45 [00:07:45] at the same time. So I think that's um, you know, really the idea of all of this. [00:07:50] And this is a lot of, you know, stuff that probably Drew and I have done for specific problem solving purposes over the course of two years with this stuff. So yeah, I think it'll be, you know, if you [00:08:00] have I'm going to put this in here, right? If you guys have some cool stuff that you're doing with the tools too, we probably can't cover everything in an hour and a half, drop them, drop them in the comments, let us know.

Play from 0:08:10 8:88:88 0:08:10 [00:08:10] This is like, this is community sharing. Want to see what everyone else is doing. It's going to spark some ideas. [00:08:15] I'm sure. Drop it in there and agree. We'll talk about it on, on another podcast just to see if there's something that we can test and have fun with too. Great point. We we're, we're learners too, right? [00:08:24] Like we, we, we're always looking for. New things, New ideas that we haven't thought of new, [00:08:30] new things to test and experiment with. So yeah, we're, we're sure you've guys got some great things as well. Share those for us. Um, in the meantime, man, you want to get us started with like number one and I don't know if we want a number of these things or what, just to kind of make it a little bit more digestible, but, um, kick us [00:08:45] off with something, yeah. So I think, look, if you're just starting Midjourney, you're having trouble prompting, right? Think you can utilize chat, GPT, Claude, anything to number one, help you expand your

Play from 0:08:59 8:88:88 0:08:59 knowledge of [00:09:00] some, some certain aspects of what we talk about a lot, which is this prompt structure, right? So drew, and I typically have a pretty solid, I think, prompt structure that's usable in a lot of different formats for a lot of different Images, right? [00:09:11] Like this can span the gamut from, you know, illustration [00:09:15] to photo realism, to, you know, 3d, whatever you're thinking about, but it's really about taking an image, breaking it down into his requisite parts. So we look at, you know, something like the photograph, right? Like we have our shot type, we have our subject in action, we have our composition, [00:09:30] we have our textures, our colors, um, you know, lighting.

Play from 0:09:33 8:88:88 0:09:33 [00:09:33] Things of that nature, If you're not familiar with certain aspects there, go utilize the tools to suggest some other ones. So, I always think about it like, Oh, composition, great, like what, how [00:09:45] can I learn more about composition? Let me go ask ChatGPT. What are some typical compositions for Cinematography, not just photography. [00:09:52] What are some typical compositions there? What is, what is made up, you know, what, what makes up those compositions? And you get something like, Oh, you know, you can look at, [00:10:00] you know, motion and balance, or you can look at something like, you know, framing and silhouettes. And it's like, okay, well, what type of techniques would you use for framing?

Play from 0:10:08 8:88:88 0:10:08 [00:10:08] Right. And it'll tell you, okay, silhouette framing or, you know, constructed framing or this framing or that framing. And then you're [00:10:15] like, well, what is that? Great. You ask it. And then either one, you can just go take that term and test it out on Midjourney or to like, try to describe it to Midjourney so that it can, you know, really boost the image. [00:10:24] So that's how a lot of the things. I started to essentially like build my [00:10:30] prompting, I don't want to call it a library, but like a knowledge base in a sense for myself. So, you know, I think that's really important because there's a lot of, again, you go into something like lines. There's vertical lines, horizontal lines, diagonal lines, [00:10:45] implied lines.

Play from 0:10:46 8:88:88 0:10:46 [00:10:46] There's like all these crazy little photo techniques that you can utilize just, just in composition. But I mean, there's also, you know, different types of, you know, photo effects that you can put on there. There's different types of lighting effects you can put in there. It's cool to learn them, but [00:11:00] of course, you know, you're not going to know every single one of them when you start. [00:11:02] So there's a good way to go and break it down. I know you've probably done this a million times the same way. It's great. I think that's the perfect number one, right? Because going back to the beginning of understanding how to best use Midjourney. You know, [00:11:15] you, you gotta sort of figure out what you don't know.

Play from 0:11:17 8:88:88 0:11:17 [00:11:17] And, uh, you know, like for me with, with Midjourney being a, an image tool, right, like, and just having some background in photography, I think that helped me get up to speed a little bit faster in terms of like, you know, camera angles, like it can, [00:11:30] it can do this, right. If I put in a, A camera or a lens, you know, especially early on, it seemed to make a little bit more of a photo realist, realistic difference, you know, and then what are all the cameras that I could put in here? [00:11:41] The lenses, the camera angles. So I think in large [00:11:45] part, especially if you're listening to this and you're still toward the beginning of your. In Midjourney, not to be too meta, you know, like this is the easiest way to really expand your vocabulary or your knowledge base, like you said, because [00:12:00] camera angles, you're not going to know what all those are.

Play from 0:12:02 8:88:88 0:12:02 [00:12:02] And you're not also not going to know which ones work. Right. So like what I would typically do is go through those different holding categories, which is, I love that you sort of coined that, but figure out what those are, right. And then figure out which ones are going to [00:12:15] work for you. Or be a little bit consistent within that, because they don't all work and some are more consistent than others. [00:12:19] Right? And so like going through that, what do I know? Um, what do I know that works? What is a term or a phrase that I'm not familiar with? Let me try that. What does this [00:12:30] do? And so you think about going that through that with photography types, camera angles, um, lines, like you mentioned, composition in general.

Play from 0:12:38 8:88:88 0:12:38 [00:12:38] Um, you mentioned framing, lighting, shadows, um, moods, textures, poses, [00:12:45] expressions, um, I mean, shit, man, styles, right? Like the list goes on and on and on. And so, you know, to really, I guess, expand that you need to, you need to play with it. And I think, um, now, right, there's so many people on Midjourney and the explore page is [00:13:00] sort of built out a little bit, a little bit more comprehensively than it used to be, because you used to not really see a that you see today.

Play from 0:13:09 8:88:88 0:13:09 [00:13:08] I mean, I think there are just a lot more people using the tool and using it well, but again, I think that's a perfect way to start, [00:13:15] you know, fill in those gaps. That's why I do it like a lot of things on those categories, those, the tokens, because I don't think a lot of people really go to sit. They're not just sitting there and thinking, imagining what tokens exist, right? [00:13:27] You want to try things that, you know, people have [00:13:30] tried and save a little bit of a time and energy in the process. It's a good point. I mean, like, You know, I think the easiest way to do it, if we're going to go be actionable about this, like pick something like lighting, right? And just go [00:13:45] through, like, go ask ChatGPT, can you suggest like, you know, 25 lighting techniques that professional photographers use?

Play from 0:13:53 8:88:88 0:13:53 [00:13:53] And then like, give me a short description for what they do, That's a, you don't even have to ask, like write a really long prompt. It'll just give it to [00:14:00] you in a list and it'll be like, you know. You know, side lighting comes in from the side, paints half the face, top lighting comes down, you know, or under lighting, under lights under the chin gives you this type of emotion, or volumetric lighting, or whatever. [00:14:12] So you can get an understanding because then that also gives you like a, oh, [00:14:15] start to build the vision of like what I can either prompt with that or how I can change.

Play from 0:14:19 8:88:88 0:14:19 The other thing for that too, if we're going into something like lighting, let's talk about that. There's, instead of it being like a technical terminology, you can also give it like a mood. [00:14:29] So maybe you need [00:14:30] some adjectives, right? Like sometimes using a good word or like an emotional word attached to lighting gives you something really great. Like Drew likes to give me some shit every once in a while. I like to use ethereal lighting. A lot, because it oftentimes has an awesome effect. [00:14:45] Um, that one, that one's good, but I've used, like, different, you know, powerful keywords there, like gossamer lighting.

Play from 0:14:52 8:88:88 0:14:52 [00:14:52] Right, it was a little softer, a little bit more. Instead of saying soft lighting, gossamer just, like, paints a different picture. You know, so, sometimes you can use it for [00:15:00] adjectives, for words like that, too. Like, give me, like, something, you know, instead of ethereal, what's another word? You know, word that we can utilize there to give the same sort of effect and emotion, but not. [00:15:10] So it'll pick that up and kind of paint the entire image in that mood as [00:15:15] well. So, you know, again, the same way you're using it to write ChatGPT in a lot of ways, you can use it for Midjourney, like, or take this prompt, like, does this, does this make sense? You know, is this like, how, how would you envision this?

Play from 0:15:28 8:88:88 0:15:28 [00:15:27] Realistically, I'm always thinking of, you [00:15:30] know, what are different words? Like what's it, the same way that I would try to write sales copy, and instead of using, you know, big, can I use the word enormous? [00:15:39] Yeah. Ways to, to punch it up a little bit. Let's call it that. I think the synonyms is a [00:15:45] great, you know, Jumping off point to the initial point because training data isn't so perfect on every item that exists in this world or every, you know, scene that exists in this world. Right. And so I think, um, when you think [00:16:00] about, like, I, I think about using this for a client.

Play from 0:16:04 8:88:88 0:16:04 [00:16:04] Particularly in like metal recycling. There were, there were a few equipment pieces that Midjourney had no idea what they were. And so like I used it to brainstorm, Okay, [00:16:15] what are like eight other names, you know, what are like several other names for this that are used globally? Tried all those. Right? [00:16:20] Because maybe one of those works and I've run into that situation a lot, right? Where it's like the first thing that comes to your mind, you would think is going to work. It doesn't, doesn't recognize that. But if you [00:16:30] change it to something else, you know, like I just had a training with a guy yesterday, actually, he was trying to put a sign on a gas station pump, but he was thinking of like, you know, like the, the digital display.

Play from 0:16:41 8:88:88 0:16:41 [00:16:41] Where it kind of shows the gallons and how much, right. But [00:16:45] instead it was giving them like a billboard on top of it, even though he was just saying sign. And I said, well, I think, you know, just try digital sign to what happened. Got it. Right. And so like you, some people assume that it's not going to get certain things in, they haven't gone through the process of trying.[00:17:00] [00:17:00] Different synonyms or adjacent ideas. So that's a really good one. The other thing you can do with it too. This is, I knew this was going to happen as soon as we start going into it. What did you got? Was, uh, was colors, right? Because I think everyone is [00:17:15] pretty much their baseline. Their baseline is, you know, red, blue, yellow, you know, whatever.

Play from 0:17:20 8:88:88 0:17:20 [00:17:20] Using, using ChatGPT or Claude or whatever to come up with different shades of those colors. Because Midjourney is actually pretty good at it. Like I found, I found one [00:17:30] shade of green that I love. Instead of just like dark green or turquoise green. Like I was trying to find something in the middle. And it came up with Phthalo. [00:17:38] P, it's like P T H A L O. I looked it up and I'm like, this is a sick color. I love this. And I started prompting with it. [00:17:45] Hit it, you know, so you can find like, you know, instead of blue, cyan or sapphire, things like that, where you can get different shades and different like emotions from those colors. I think that's a really fun one to play with because you get that's a, that's an important one to, you know, [00:18:00] because also at least up until now, hex codes, Pantone's, they're not specific.

Play from 0:18:06 8:88:88 0:18:06 [00:18:05] They Midjourney doesn't respond to those one for one. So. I don't, I don't think there is anything one for one, but I did have something interesting happening the other day. I use that suggested [00:18:15] prompt when I was doing like a external editor or retexture or whatever I was doing. And, um, it gave me a Sherwin Williams color and I was just like, interesting. [00:18:27] Eggshell [00:18:30] white. Like, man, it was like Sherwin Williams color number this.

Play from 0:18:35 8:88:88 0:18:35 And I was like, Oh, Oh my God, you know, and I haven't, I haven't, uh, I haven't experimented with that, but [00:18:45] I just might. Yeah, I think that could be interesting, but yeah, to your point, it does pretty good with, with most of the, the names, uh, you know, like lime green, um, you know, [00:19:00] things like, you know, like just different variations, canary yellow, like anything like that, uh, it's pretty solid because it makes a difference, right?

Play from 0:19:08 8:88:88 0:19:08 [00:19:07] Like when you see, when you prompt red, most of the time it comes out red, right? Like it looks like whatever the most. You know amorphous sort of [00:19:15] collection of red is it's not, you know, like a deep like warm red It's not like it's like a cherry red most of the time. Yeah prompt red So, you know, there's how many different variations of red are there like a billion so you can get in there and like [00:19:30] Describe it a little bit more and I think that time sometimes is, you know, again, even utilizing different analogies there as well, like ladybug red, something like that, right?

Play from 0:19:39 8:88:88 0:19:39 [00:19:38] Like that's a, that's a little bit different than a rose red or, you know, we're getting micro here, but again, that's where, [00:19:45] what do you want the control and you want to play with the tools. You can see that it does have power like that. So yeah, it's fun. That's a good way to start. Yeah. I love that. Dude, that's a, that's a, that's a good one. [00:19:55] That's a really good one. Where are you, where are you taking us from here? Yeah, [00:20:00] let's go like more of like the story route, you know, to, uh, I think I know you've got some things that at least come to mind for me that you've done. But, uh, you know, with, with this, I mean, more or less like, Let's say you want to come up [00:20:15] with a sequence of images, you know, to tell a story, you know, you want to do a story arc, you know, obviously you can use ChatGPT to come up with stories.

Play from 0:20:24 8:88:88 0:20:24 [00:20:24] You can provide them with certain frameworks of stories, right? Because there are different types of frameworks, [00:20:30] um, that are used in stories, but there are like some general ones that you just, Hey, like, I've got an idea for this minimal effort. This is the kind of story arc I wanted to follow right out the story. [00:20:41] And then leverage that story to say, okay, based on [00:20:45] this, if I were to create X number of images, what images could I create that would tell this story? Completely, right? Just using static images and then can you come up with the props for each of those, right? Using X structure, right? So you're basically using that as a [00:21:00] as a complete partner all the way through the story telling process.

Play from 0:21:03 8:88:88 0:21:03 [00:21:03] I've done that several occasions. The thing that I thought also thought about for you was, uh, I know you did this early on, like one example that still stands out for me was the one that you did [00:21:15] with Tulum, right? Or you, you like, basically like tell people about that because I thought that was fascinating. [00:21:20] And this is a long time ago too. So like, no one was really doing this stuff yet, but you basically utilize ChatGPT in a way to kind of like, [00:21:30] I don't know, how'd you go about that? You basically created different scenes, right? As if you were telling a story of somebody traveling to a hotel. Yeah, so this was when I, when I realized that we could operationalize prompting and when, you know, everyone was before SRF, before you [00:21:45] could really do anything with, uh, you know, Image and style consistency.

0:21:15Story Building

Play from 0:21:49 8:88:88 0:21:49 [00:21:49] It was like, all right, if I can't like reference something, I'm going to have to, I can't use a Laura, I can't use any of this stuff. Like I have to, my text prompts have to be so dialed in that there's no [00:22:00] difference in between the images, right? The aesthetic, the mood, like I have to really break it down. [00:22:03] That's where the prompt structure started with. So, okay. So we have to take like, what is an image, break it down into its requisite parts, and then whatever those, that branding of the images [00:22:15] that has to be. Okay. Consistent. So the textures, the colors, the composition, the mood, the lighting, the shadows, all that stuff has to be consistent.

Play from 0:22:25 8:88:88 0:22:25 [00:22:25] And then we can just flip the subject and the environment. I've really, that should be, that should [00:22:30] be the case. So what I did with, with that was I just basically started to ask ChatGPT. I'm like, if I'm looking for these specific parts of a, of a prompt, right? Like I want this. But I'm trying to walk through a hotel. [00:22:42] So in my mind, I didn't have the good, I didn't have like the [00:22:45] storyboard prompt down yet, but I'm like, how would I walk through a luxury hotel? I'd go in the front entrance. I'd go to the front desk. I'd probably get a welcome drink or something like that. Then I'd go and see the grounds. Like, where do you go first?

Play from 0:22:57 8:88:88 0:22:57 [00:22:57] The, you know, go see the pool, go see the bar, go to [00:23:00] your room, come back, go to dinner. So I was like trying to weave my way through that story, but I actually have, let me pull this up. I actually have the prompt I used for that. Cause it still hasn't changed. So boom, this was all right. So [00:23:15] when I was doing this. Right. I was using like a two part prompt. Basically I would say like, this is a Midjourney prompt formula, right in the prompt. I'd give it the prompt formula, which was like photo type subject and action environment, color scheme, camera, lens, film, type, lighting, [00:23:30] composition, additional details, parameters.

Play from 0:23:32 8:88:88 0:23:32 [00:23:31] Right. So I put that in there. Respond. Yes. If you understand, cause I didn't want Chad, JBT to go and like do it's going crazy thing. It's much better now than it was then. So like, let's create three prompts for a five star [00:23:45] luxury hotel in Tulum inspired by tree houses, right? Like that's what it was back then. [00:23:50] Use concise, visceral, powerful language. Only use keywords, do not use filler words. So then it would just go and rip, like here's three [00:24:00] prompts and I'd go, okay, let me go try those out. So then I'd go test them and I was like, okay, cool. Now, if I said like, all right, now in the same thing, I wouldn't change anything else right with subject and details, I would just say like, okay, for this five star luxury hotel, you know, uh, pool [00:24:15] shot, like a shot of the pool, like instead of just like changing the entire prompt, do everything the same and then just say,

Play from 0:24:20 8:88:88 0:24:20 you know, luxury five star hotel inspired by, you know, tree houses, view of the pool. [00:24:24] Okay. Or the lobby or, the bar or a hotel suite X, Y, [00:24:30] Z. So then I realized all it didn't, didn't have to change that much. And I could just do everything really consistently. [00:24:34] So I redid this a while, like, I don't know, maybe a couple of months ago. But it was essentially like, how can I redo what I did in Midjourney version? I think it was, I don't know if that was four, [00:24:45] maybe that was five. But it was again, just getting these different shots, right? Like it's just the same prompt over and over again.

Play from 0:24:50 8:88:88 0:24:50 [00:24:50] I'm just changing like the room. Yeah. Because it all looks, you know, the same. Then you can use this in the same style to get the food to make it all look consistent. Now we can use style reference. [00:25:00] Right. But again, you can go through and have like this entire story, this entire branded story with basically text prompts if you want it to. [00:25:07] So I know I'm scrolling through this fast, but there's so many images and I was like, I don't even know how to curate these because I like them all so much again. [00:25:15] Um, but I think this was v6. So that was probably like, you know, two, three months ago. But that was, that was like the breakthrough in terms of, okay, there we can get consistency.

Play from 0:25:26 8:88:88 0:25:26 [00:25:26] That that is a myth that was something with version five when I [00:25:30] did it does it so everyone knows those yeah images we just saw were updated I did well you know that reminds me to like where we were at in time that there weren't um there wasn't a lot of talk about problem structure yet but I'm trying to think about what the impetus or the inflection point was for [00:25:45] that and I think in a lot of ways it was trying to get consistency When, whenever you didn't have things like style reference at the time.

Play from 0:25:53 8:88:88 0:25:53 [00:25:51] Mm-Hmm. . Right. All you had was, was image reference. I can't remember how good it was at that time. You know, it was okay, but, but you [00:26:00] know, at the same time, right. With no color, you know, identifiers, things like that. Like it was really just about keeping something static and changing. pieces of the prompts, you know, so having that template to create that story, [00:26:15] you know, and so I think that was sort of how these prompt structures came to be.

Play from 0:26:19 8:88:88 0:26:19 [00:26:19] Midjourney never came out and said, Hey, use the same prompt structure every time. That was just something that people came up with because they found out that they could use, you know, they could just get better consistency in doing so. And so I thought [00:26:30] that was like a, sort of like one of those, another one of those like cool little Easter eggs where. [00:26:35] Midjourney didn't do that on purpose, but like something really cool came out of it because people were creating that was also You know, i'm patting myself on the back [00:26:45] everyone. Sorry doing it once everyone every once in a while it was like when I sent you the when they updated their website and they had like The prom structure, like word for word from like my document and some of the stuff from your documents where it was like, Oh, they just lifted this straight up and put it on the [00:27:00] website.

Play from 0:27:00 8:88:88 0:27:00 [00:27:00] I was like, Oh, all right. Cause it even had like the, I mean, word for word, it was word for word. I remember that. Like, I was like, this is, you know, it went beyond just the prom structure. There was other things in there and I was like, all right. Regardless, you guys, you know, you guys have given me a platform, so I don't really care.[00:27:15] [00:27:15] For sure. But just funny that they, you know, I don't even think they, you know, They think about a lot of this stuff themselves, just like, here's the tool, go create, but they also need someone to explain it a little bit. I love that, you know, and also like another cool, cool idea too, is just prompt prompt [00:27:30] variations, if you get a prompt from chat, GPT, or you've got a prompt, take that in and get variations of it.

Play from 0:27:35 8:88:88 0:27:35 Test those against each other, right? Like again, uh, there may not be much difference in the prompt, but you're probably going to get two totally different results, you know? So depending on what you're using it, uh, you know, [00:27:45] prompting for, right? [00:27:45] Like maybe if you're just doing personal stuff, you don't care so much, but if you're trying to achieve something specific and, you know, kind of find like those best images, that's another way to, to dig. Cause you can also expand that prompt, right? Like I do do that as like the [00:28:00] baseline, like here's the key powerful terms that I want in there.

Play from 0:28:03 8:88:88 0:28:03 [00:28:03] Like we're prompting version 5. 2 again. Right. Yeah. Yeah. And then it's like, Oh, expand on this. Give me more detail. Right. Like, uh, that's really where, you know, keep it under a hundred, whatever the, whatever the token count is or the [00:28:15] character count is that you can put in. But you know, it used to have like these paragraph prompts I used to see from people. [00:28:20] I'm like, Oh, that was generated by ChatGPT. Um, you know, we did it too, but again, like you never know what, when you write that, the problem here is either iteration, right? So when you [00:28:30] put, we'll talk about like the good and the bad of this too, is when you get like these really long paragraphs. You never know what, what, if you don't like something in the image, like, where is it, where is it in this two paragraphs that like screwed up everything that I don't like, because [00:28:45] then it's like, there's so many variables.

0:28:30Prompt Structure

Play from 0:28:47 8:88:88 0:28:47 [00:28:47] You don't know what's affecting the image. And it's like, now I have to go change like everything and rewrite all these sentences. So sometimes it's easier to start out with just those keywords and then expand. Yes, you know, I think that's like a good way to use the tool also, [00:29:00] but yeah, I like that though that, I mean, so that's, that's part of it. [00:29:03] Right? Like the prompt variation. And I think like just you could have a structure. There's also like times I found like that, you know, uh, Midjourney can actually, you know, do a lot more with the, [00:29:15] With longer phrases than I thought that was sort of like, uh, part of the thing I experienced when doing that custom GPT for like reverse image generation kind of, um, it was just more or less like, Hey, I don't really care about the structure.

Play from 0:29:27 8:88:88 0:29:27 [00:29:27] I just care about you. Like [00:29:30] describing this in full in totality, right? Like the hat, you know, this guy's wearing a hat. It's black What's the hat made out of right? Um What's on the hat right like just? Describing it fully [00:29:45] as if somebody cannot see it and just getting like really long Um, prompts out of it and then just copy and paste in those. [00:29:50] And that worked pretty well too. Like, I'll just share a quick example of that because that was sort of the idea I built, um, when trying to do with the, with the GPT is it's [00:30:00] like, I just want to be able to see if I can crack this code. Um, let me go over here and get into this so you can see this, right?

Play from 0:30:11 8:88:88 0:30:11 [00:30:11] Okay, cool. Like, you know, so you just take any image, right? [00:30:15] Um, And then, basically, it's giving me paragraphs, and these aren't even, I probably have to go back pretty far, uh, I haven't, I haven't really been using this in a minute, um, worth of, uh, pages worth of prompts, yeah, yeah, yeah, like, man, [00:30:30] these aren't even, yeah, I mean, but basically, right, like, it would give me, it would spit out some long, long stuff, yeah, right, and, and, honestly, I would, it would interpret it really well.

Play from 0:30:41 8:88:88 0:30:41 [00:30:41] The coherence was better than I thought. And then, you know, even if it's not, you can then [00:30:45] always layer on the image reference on top of that. Right. So like, I just found that to be a great way to use ChatGPT to skirt around the default described tool that Midjourney had, because, you know, the described tool was really cool to use, but the one knock on it [00:31:00] was you're not going to get. [00:31:01] Quite what you're hoping for out of it, right? You can still get some cool generations, but it's probably not what you're hoping for. So I thought that was like, yeah, you know, like it was close, but it wasn't like exact and that's where these tools come in. And so like, I love using the vision [00:31:15] aspect. I think that's the underutilized part.

Play from 0:31:18 8:88:88 0:31:18 [00:31:17] Another example, right? Like I, you know, I went on that kick where I'm just. Pumping out all these style codes and look, I don't know every single style that exists and how to describe it, but hey, look, like all [00:31:30] I did was like, let me go all the way back to the top. How would you describe the style of image? [00:31:34] Yeah. Boom. Right? Like, so now I can describe it to people. Right. And then it kind of puts them into their own categories. And so I was really just describing these of like, look, this is one that's, you know, [00:31:45] surrealistic portraiture with this kind of aesthetic. Right. And so, I mean, like, the vision aspect, I think, is so, so underutilized.

Play from 0:31:52 8:88:88 0:31:52 [00:31:52] There's also the component of you could take an image and have it assess. The image for like how [00:32:00] compelling it is, whether it's set up compositionally correct, like, like graded on a scale of this to this in terms of like how great this image is, what can improve on this image, right? Like you could use that as a creative partner and expert and essentially [00:32:15] audit your images. [00:32:16] Visual heat mapping as well. Visual heat mapping. So the visual part, man, dude, visual is, you know, now that we're just kind of like talking about how to describe that. I think that's. That's the underutilized part. The [00:32:30] visual components. Cause it's, you know, again, like the heat mapping stuff, if you're in marketing, you want to know where people, people's eyes are going.

Play from 0:32:36 8:88:88 0:32:36 [00:32:36] You know, like that's a, that's a big part of like, you know, you don't want someone's eyes going to the upper right hand corner if what you're trying to do is get them to click on the bottom [00:32:45] left hand corner. Right? Like maybe you do. I don't know. But most of the time, that's how I'm thinking about it. [00:32:49] It's like, where can I put the focal point of the image so that it still is like within eye tracking range of, you know, wherever, whatever I'd need them to do, which is click here. Right? So like, that's, [00:33:00] that's where it can be helpful. It can also just, you know, again, which one, which one catches your eye?

Play from 0:33:05 8:88:88 0:33:05 [00:33:05] Like I like to ask it, you know, which one catches your eye and this for this audience, like for if I'm scrolling on social media, you know, even if the, the answer is [00:33:15] the one that I don't like. You know, sometimes that one performs better. Um, because it's just like, it's, it's actually like, you know, the colors are actually slightly more saturated. [00:33:23] And there's a little bit more contrast near the focal point of the image where X, Y, Z, blah, blah, blah, blah, blah. It's a little bit more objective. It's a, it's an objective [00:33:30] thought partner, which, you know, a lot of people have opinions, right? And we have opinions, but to your point, like you might like an image, but it's hitting you with a couple of facts.

Play from 0:33:39 8:88:88 0:33:39 [00:33:38] So it's like, okay. And so you could do that with the Midjourney image straight out. You could also do that with a, [00:33:45] Design in general. Now you've layered in text on top, right? Like I, I did that recently too, where, um, you know, I put together an image, then I started to do some graphic design on top of it, some elements, some text. [00:33:58] How would you grade this on a scale of [00:34:00] one to a hundred, a hundred being perfect, right? Like setting up some of the things that it's looking for, you know, that, that you're hoping to achieve with it. So, you know, like at the end of the day too, You do need to kind of know how to prompt and be a little bit articulate with, with the context [00:34:15] behind what you're prompting for the goal that you want to get out of it.

Play from 0:34:19 8:88:88 0:34:19 [00:34:18] And then, you know, if you're not sure about that one, one quick hack that you can always do is say, what, what else can I tell you to make your response even more accurate or more high [00:34:30] quality? You know, that's, that's another great one too, because you know, we, No one's an expert prompter, I would say. You know, like there's always more that you could add on. [00:34:38] So, you know, that's, that's to me is the exciting stuff. It's, it's always cool.

Play from 0:34:44 8:88:88 0:34:44 I love, I [00:34:45] love asking it like, really random questions too. Like, think about this, like, uh, you know, find something a human wouldn't or like go, go into non-traditional route to, to help me with this. Or do this in the exact opposite way that I'm telling you to do. [00:34:58] Right? Like, you know, go [00:35:00] run wild. Don't do anything, you know, don't do anything that you don't want to do. Just like put your own creative spin on it. Like sometimes you get some really crazy stuff when you let it go, when you let it cook. Yeah. Or I'm like, you know, take this idea. And either expand on [00:35:15] it or make it 5x crazier.

0:34:45Color Theory

Play from 0:35:17 8:88:88 0:35:17 [00:35:17] I like that one. Like, here's a prompt that I have to make it 5x crazier. Or 5x weirder. Or whatever. Um, you know, amplify subtle details of it. It's very, you know, sometimes it's fun to play around with that. [00:35:30] Um, the other thing I have here is, actually, could start, could do this too. I pulled a couple of these prompts because I just, I thought they'd be easy enough to try if anyone wants to try this one. [00:35:39] Like let's see. What do we got here? Thought partner. Cause he said thought partner. You reminded me of [00:35:45] it, right? I used to give this to people who were just starting or just like, really, you know, this is in, in portions of, I think it's in my course somewhere along the line, but I'm just like, if you don't know what to prompt to, and you just want to play and you're feeling creatively dumb, like I get this every once in [00:36:00] a while, because I know what I'm trying to get out of it.

Play from 0:36:02 8:88:88 0:36:02 [00:36:02] And sometimes I just need some inspiration, but I just have it as a thought partner, like you're not my creative partner. Let's develop some awesome ideas. I don't need prompts. I need you to be a thought partner, right? Like I like to use the three use, which [00:36:15] is, this is just, I made this up because I thought it sounded cool. [00:36:17] Like unique person, unique place, unique circumstance as a framework. Interview me as if you were my enthusiastic, creative partner, and then just let it take you through the process. And it'll give you something Like really substantial, you [00:36:30] know, I think that was that was a good one for me to start would get people like rolling like okay You know, like, uh, you know, you're like, what do you like, you know, I like, you know travel photography stuff like that I'm like cool, you know unique person unique place unique circumstance And that's how you end up with like, you [00:36:45] know, a dude herding sheep in the Rainbow Mountains of Peru, right?

Play from 0:36:49 8:88:88 0:36:49 [00:36:48] Like it's it ends up getting you some cool stuff. Yeah, because you're not thinking about that Most of the time, if you're not inspired by it, but that reminded me of like almost the idea of like, uh, [00:37:00] what do I want to call this? Like thematic expansion. Yeah, it sounds, I don't know, but like to your, like, if I want, if I've got a certain style that I go for, let's just call it, like, I like to use a lot of juxtaposition. [00:37:13] Yeah. Right. And then I [00:37:15] say, Hey, I like to use, you know, I like to create juxtaposing images that do blank, blank and blank. Here's an example of what that is. So you, you know, maybe attach an image, show it, come up with 25, 50 more ideas, [00:37:30] similar to this one, right? So, I mean, like it's just that part of it, the brainstorming part of it is so great,

Play from 0:37:38 8:88:88 0:37:38 you know, because it's like, all you really need is one thing and then you can build, build, build, And you can even take, you know, [00:37:45] go from one to zero to one. You know, with, with ChatGPT. So that's where it's, it's so versatile, right? It could be something that you use at the beginning, the middle toward the end throughout, right? And, and [00:38:00] so [00:38:00] that's a really solid point. I didn't mean to cut you off. I just wanted to, I just know you didn't, you didn't cut me off. That, that was good. That was thanks, bro. That was solid. I like, thanks man. Yeah. Well, I'll give you a little bit of a little credit there. I mean, you got my mind going, so this is what [00:38:15] I'm saying.

Play from 0:38:15 8:88:88 0:38:15 [00:38:15] I knew, I knew once we get started with this, it's like, damn it. This could be like six hours. Cause you forget how much you use it for certain things. Like the other thing too, is, you know, I have, maybe, maybe we'll do it now. Maybe I'll save it towards the end. Like from a prompt structure standpoint, when you're trying to get like ultimate control [00:38:30] or extra pop, I talk about, like, these prompt formulas, right? Like, these prompt formulas, like, aren't set in stone. It's not the same prompt formula you have to use every single time. It's just, like, the baseline. Like, these are covering the controllable aspects of an image. Now, [00:38:45] you can expand upon that, which is what I like to do a lot of times for specific imagery.

Play from 0:38:50 8:88:88 0:38:50 [00:38:50] So, you know what? I'm just gonna jump into it. Fuck it. Um, I'll just show this one. Because it's too hard to explain without showing it. We're gonna go live. Fuck it. this is one of my favorite ones to [00:39:00] play with. Now everyone likes to utilize, you know, your style reference, your image reference, whatever. [00:39:05] I'm still a text prompting guy. Cause I feel like that's the most controllable aspect of it at this point, um, to like really get the pop and then play with style reference after and things like that, but [00:39:15] sometimes I want, I want to like be very specific in what I'm trying to get or when I'm working with clients and they really need something that's very specific.

Play from 0:39:22 8:88:88 0:39:22 [00:39:22] I'll go into, I call it my like little prompt formula builder. So it's not just like that standard prompt formula. It's what are like [00:39:30] very specific categories that are like relevant to each photography theme. So you think about food photography, right? Like you don't really care about most of the time, like the architecture. [00:39:42] In the room, you care about like the, you might have [00:39:45] like the composition, the lighting, the colors, X, Y, Z. But what, what's the plating, what's the garnish, like what, what type of other descriptors can we have in there about the food to control the food more, right? Like it's a little bit more focused than just like [00:40:00] general photography.

Play from 0:40:01 8:88:88 0:40:01 [00:40:01] Right. So I used to, I use this a lot, like, your generative prompt engineer, our goal is to take photos and break them into prompt categories. Categories are key elements that would make up an award winning photo. We already understand the, the standard categories of image composition, [00:40:15] you know, lighting, photo type subject, things like that. [00:40:18] So suggest unique categories. We could, you know, Uh, focus on to be ultra detailed with theme. So it's like, it's like, you know, food or architecture photography, right? Let's just using those in an example. We're looking for categories like, [00:40:30] and then I would say, you know, for architecture, like interior design style or fabrics or textures or patterns or things like that.

Play from 0:40:38 8:88:88 0:40:38 [00:40:37] So it knows what I'm talking about. Cause then it would give us examples like this. Like these are some of the categories you might want to look at. Right [00:40:45] for being a little bit more detailed in architecture, architecture, character, space, and flow landscape surroundings, material quality, and finishes, you know, Vista's atmospheric conditions, lifestyle, emotional appeal, detail, and craftsmanship. [00:40:59] Cause when you start to [00:41:00] then build out a structure. Like a prompt structure or something, how you want to fill it in just to like, keep this stuff easy for you. Then you have like your standard categories, photo type subject, um, you know, environment, lighting, mood, technical details, but you can fill it in with, you know, these other [00:41:15] pieces like room environment, material finishes, color scheme, interior design.

Play from 0:41:19 8:88:88 0:41:19 [00:41:19] And then, you know, you can have either yourself fill it in and you just have chat, GPT, fill it in for you. It's like, once you structured this prompt, which again, you can have chat, GPT, do it for you, or. Trying to create like a living room [00:41:30] inspired by, you know, the, the caves in Cappadocia in Turkey, right? [00:41:34] So it just comes up and fills all this in with open circular layout around a central fireplace, handwoven rugs, low wooden tables, plush cushions, like all these little things make a difference. [00:41:45] Cause when, you know, you put it all together, you get what you're looking for. Right? Wait, that's really something.

0:41:45Visual Assessment

Play from 0:41:51 8:88:88 0:41:51 [00:41:50] I like to do to push the envelope. You can do this. It works very well with food photography and the controllables. But like, again, I was just using this living room as an example because you might [00:42:00] need a little bit more firepower for being ultra specific. Um, that makes sense though. What was that two out there? [00:42:07] No, you know, you think about a car photography, right? Like totally different than food photography, street photography. Yeah. Different than [00:42:15] to your point. Like we, we covered that in the tokens where it's like that, that holding category can do some heavy lifting for you because it, it really understands what those things are, you know, and then can really work within those constraints, which is probably going to leave you [00:42:30] to more consistent and accurate results.

Play from 0:42:33 8:88:88 0:42:33 [00:42:33] And then again, once you have to me, once you've broken those categories out like that iteration to me is like, cause how many times do you prompt? And it's like, what? First shot. That's it. I'm done. Like, you [00:42:45] know, most of the time you're playing with it a little bit, right? It's just easy. Once you have the categories, by the way, that doesn't, that doesn't happen ever, ever. [00:42:55] There's, there's certain times it's weird. I get it. Like when I'm, when I'm doing like a runway or like, [00:43:00] uh, you know, uh, AI happens more now than it used to though.

Play from 0:43:03 8:88:88 0:43:03 Yeah. Yeah, I'll just like oh, whatever I can work with that, you know, like instead of trying to be perfect for like the yeah It's gonna be presented But like I just love having things broken out like that via text because then I know [00:43:15] what's screwed up So it's like okay the lighting. [00:43:17] I don't like the lighting. Let me just go to the lighting category change the lighting go from there Yeah, instead of trying to like dig through this entire rewrite a prompt Cause I didn't like the lighting. It's just like, you know, find, execute, iterate, find, execute, [00:43:30] iterate, like pretty simple. that reminds me of like, we, we've got a lot of people, different types of people that listen to this podcast, but also that use Midjourney.

Play from 0:43:40 8:88:88 0:43:40 [00:43:40] And so, you know, I'm gonna, I'm gonna give you two more ideas here. Just thinking back to [00:43:45] previous episodes. One was a side hustle in use case ideas, Like if you're trying to come up with different ways that you can use AI image generations in your role at work or for personal gain and side [00:44:00] hustles or build a business or make money off of it. [00:44:02] Great brainstorm partner for that stuff. Like huge, right? Like, so if I'm like, I'm in marketing B2B, what are all the different things I can be thinking about how I could use these images? Right. I mean, again, could I go through and name a handful of them? [00:44:15] Yeah. But like, do I have a full vision of what's possible?

Play from 0:44:17 8:88:88 0:44:17 [00:44:17] No. So that like, that to me is a great idea. I think also. you know, because people find themselves in different roles, those things are used differently. So for example, an architect is going to use like, or, [00:44:30] you know, somebody that's an interior designer is going to have different use cases for this, right. [00:44:35] Which then could impact everything else that goes on top of that, of how you could use ChatGPT in combination with Midjourney. the other thing that like same thing, [00:44:45] yeah, fashion finishes and textures and patterns and all that other stuff, right? Like there's a lot of different things that make up.

Play from 0:44:51 8:88:88 0:44:51 [00:44:51] Fashion is not just like woman in a dress. It's like the, you know, got to go into details about the dress, how it fits, what the patterns are, you know, how it sits [00:45:00] on the body, like where the accentuating side, you know, pieces of it are like, that's, you know, what's the stitching like, like, that's all like little micro categories that you can control. [00:45:09] You just have to sometimes, you know, break it down that way. So it's, it's makes sense for [00:45:15] every thematic, like a image set that you're trying to create. It's like, how deep do you want to go? Cause, cause you can go deep if you want it. Here's, here's the other thing about what I've noticed in using large language [00:45:30] bottles since they've been released, the more you use it.

Play from 0:45:33 8:88:88 0:45:33 [00:45:32] The more you think of other applications for it, so you're going to start to like naturally and organically find more ways to use it and use it more often. So if you can kind of like break through that, you're going to [00:45:45] discover your own ways and processes too. I'm more or less speaking to, I guess, people that don't use this stuff so much, but that's, that's one big thing that was an unlock for me early. [00:45:54] It was just like, okay, now that I'm using this for this, I'm, I'm thinking, Organically about all these [00:46:00] other ideas now that I wouldn't have thought about it before, you know, and then so you'll like, bring it up to somebody like, I never would have never thought about that. It's like, well, I mean, it's just like when you use it every day, you start to think of the categories that you use it for, right?

Play from 0:46:12 8:88:88 0:46:12 [00:46:12] Like ideation, iteration, [00:46:15] visual assessment, right? Like all these things. And so when you have those problems, you think of it. I, I also have a really nerdy technical use of it, which is like a lifesaver cause you know, designers out there. You [00:46:30] might love this, you might hate it, I don't know. But whenever I have like a collection of images, right, that I want to format into one page, or I want to format into one design, oftentimes they have different sizes, different visuals, or different, uh, aspect [00:46:45] ratios.

Play from 0:46:46 8:88:88 0:46:46 [00:46:45] Like a lot of times, like, uh, for, for example, right, I'll use, I'll use, uh, the videos that I create on, on LinkedIn, or on Twitter, right, where I'll have, if I'm trying to show like, Runway, right? Like a visual of runway where I [00:47:00] have a key to key frames and then like a video, but I want the video to be big and I want the key frames to be. [00:47:05] Um, you know, smaller in size, but the same width as the image below. So it all fits into like a perfect little square. Like I've, I asked ChatGPT to help me from like a [00:47:15] pixel and like, framing perspective. So it'll be like, Oh, here's my images. Here's the baseline pixels. There are, you know, 1200 by seven 86, whatever the hell it is.

0:47:15Technical Tips

Play from 0:47:25 8:88:88 0:47:25 [00:47:25] And I'm like, I want to put. Two side by side at 50 percent and I want to put one at [00:47:30] 100 percent but I want to stack them on top of each other vertically and I want it to fit into this aspect ratio. Like tell me what the percentages need to be, the pixel sizes, all of that so that I can just do it instead of like trying to format this crap. [00:47:41] Because like I used to do that with designs all the time, you know, when I was [00:47:45] doing decks and, you know, little graphics and ads and things like that. It's like, Oh, I have six pictures. I need to fit them into a 19, you know, 10 80 by 1920 square.

Play from 0:47:54 8:88:88 0:47:54 How does it fit? Yeah, that's a good one. Yeah. With like, you know, a 50, 50 pixel, uh, space in between each of [00:48:00] them with a, you know, 40 pixel bleed on the outside. [00:48:03] It's math. So it's like, how can these things help me with the math problem? So I don't have to sit there and just play with it all freaking day. Like that's a, that's like a really technical use case, but I use it all the time for stuff like that. [00:48:15] Because I just, I can't do that. I love that. On the inside. [00:48:19] It kills me on the inside to format that stuff. I love that one, dude. Yeah. Just brought it up a lot of pain behind behind these eyes and behind my voice in that one If anyone can't tell that's years of [00:48:30] just misery trying to figure that stuff out, but it's one of my favorites Yeah, remember when we were doing uh, we were talking about the upscaler creative and subtle, right?

Play from 0:48:41 8:88:88 0:48:41 [00:48:40] Like that was another visual use where just put them side by side. What's the difference here, you know, [00:48:45] there's a lot of things that Like curation is a big part of this now picking the right things to show or use in projects, whatever it is, whatever that is, whatever that means to you. But, um, you know, kind of going back to that, right? [00:48:57] Like just an objective decider, you know, like, what [00:49:00] do you see here? Versus here, compare and contrast. Um, I also think you brought up a good one, which is like also one of my favorite things to use, which is around like unique. I like to use like unconventional. Love unconventional. Yeah. [00:49:15] Um, you know, but like get it break out of that normal box of what you're typically going to get back.

Play from 0:49:21 8:88:88 0:49:21 [00:49:20] And so that can apply to basically everything we brought up. But at the same time, like that's a really good. I think way to break out if you're not trying to like [00:49:30] go down this huge mega prompt writing spree, right? Like you can use unconventional creative ideas or unique, you know, unconventional ideas, blah, blah, blah, blah, blah. [00:49:39] there was one time I was, I was going to do it for a piece of content. This was a while back. I'm just [00:49:45] remembering it now. Cause he just sprung my sprung my imagination on this one. But I was, I was basically like, Having an unspoken battle between ChatGPT and Midjourney and I was just like screwing around one night because I wanted to document it But I never ended up making up making the content because I just probably forgot [00:50:00] about it But it was it was like hey, we got a you know, talking to ChatGPT like it's my bro like bro.

Play from 0:50:06 8:88:88 0:50:06 [00:50:06] We got a We got a real challenge here. I want you, like you're now in competition with Midjourney. Like I want you to challenge it to [00:50:15] produce the best stuff possible. Give me some like really challenging prompts and I'm going to give you the results and then you can shit on it. Like that was kind of like the idea. [00:50:22] Damn. Love that. It started, it started coming up with some really wild stuff and it was like, it was talking, you know, this is when, this [00:50:30] is where we get into like this whole meta conversation where like, Midjourney doesn't feel like AI to me, but ChatGPT does. Like where it talks back to you and like has thoughts and Midjourney just like executes, you know, that's right.

Play from 0:50:42 8:88:88 0:50:42 [00:50:41] That's a response. No, that's a good frame. Yeah. Yeah. No [00:50:45] backtalk. Yeah. And it was just like, Oh, It produced really bad. It didn't hit on any of these things and blah, blah, blah. Try this one. I bet it can't do that. I was like, all right, you took the competition to heart, right? I guess it just started like acting in that frame of [00:51:00] like, this is competition and I'm supposed to be better than it. [00:51:02] Um, but it was, it was really interesting to watch the interaction. Cause it was, it was prompting some very like, Oh, there's a woman driving in a car. It wasn't like that. It was some really out there stuff. Um, way more like. [00:51:15] ethereal and ghastly, like not dark, but like way more, uh, I don't even know the way to put it.

Play from 0:51:21 8:88:88 0:51:21 [00:51:20] It was very like fluid and not technical. It's, it was very interesting just to watch the tool, like come up with its stuff on its own. That wasn't like, no, I was like, no [00:51:30] direction. I don't want any direction. I'm like, yeah, which now we might have to do that again. Like we might have to, might have to bring that thing back to see you, you know, see how it, how it's changed since like GPT four. [00:51:39] To whatever the hell we're on now, uh, you know, 3. 5 and Sonic, it's getting, it's getting insane [00:51:45] now. Yeah, I don't even know anymore, but have you ever, you ever done anything like that? Just like, get random. Yeah. Like, I, I mean, uh, I don't know that I've like thought about like the competition thing, but I, I love that.

Play from 0:51:58 8:88:88 0:51:58 [00:51:57] I've been using it more or less to, [00:52:00] to help me write Even better prompts lately, like especially like, so this was more work related, but I was trying to come up with prompts relative to an audience, um, for a client. So it [00:52:15] was more or less like for sales leaders. And I wanted to think about how I could provide value. [00:52:19] And of course you got these people out there that, you know, Hey, 50 prompts for sales leader, blah, blah, blah. But here, here's the thing. 90 percent of them are one or two sentences. You're not going to get great [00:52:30] stuff with one or two sentences. You just won't, um, for, for 99 percent of the use cases, because they require more context behind the goal.

0:52:30Brand Integration

Play from 0:52:38 8:88:88 0:52:38 [00:52:38] What's your output is the format of it, blah, blah, blah, blah, blah. So the more ammo that you give it, the better. So what I would typically do is [00:52:45] like, even in this example, write out what I'm thinking. have it write a better prompt on top of that, you know, and then also say, well, now that you did this, like, I also think that if I provided you with A, B, and C documents [00:53:00] that would improve these prompts even better. [00:53:01] Wouldn't it? Oh yeah. Like for sure. So, so like, long story short, I was building out these mega prompts for sales leaders, thinking about the jobs to be done approach in terms of how they're going to be become more efficient at their job, more successful in their role. [00:53:15] Right. So what's, what that means is, you know, hitting quotas, you know, shortening sales cycles, winning more deals.

Play from 0:53:21 8:88:88 0:53:21 [00:53:20] Right. And so using that framework, having a build on top and then asking it, well, if I provided you with. Documents. Would that be even better? [00:53:30] Yeah. Well, what documents should I provide you with? Right? And so now all of a sudden you've got just this one big massive thing that you can take and it's so fucking valuable, right? [00:53:39] Because the, I think the hardest part is the attention span when prompting. [00:53:45] Everybody knows it should be kind of like a conversation, but you do get a little bit lazy and impulsive about just like hitting enter. You're just like, okay, I feel like I've done enough at this point, you know? And so like, You know, asking it to build on and improve and [00:54:00] suggest better prompts for better quality, more accuracy, more consistency, uh, has been like a game changer too.

Play from 0:54:09 8:88:88 0:54:09 [00:54:08] And again, like that can apply here, but also just outside the scope of AI image generation, but it's a, I don't know if you've [00:54:15] gone through the same sort of progression. I felt like when I started, it was very short prompts. And then I went and like, I got to do mega prompts and put all this context in. I didn't like. [00:54:24] I backed off that and it's back down to like, I'm just having like a legit conversation, like I'm texting somebody, you know, [00:54:30] like the memory has gotten so much better where like, kind of understands what I want half the time now already without having to ask for context. I send people my conversations a lot, you know, cause they're like, they're just like, dude, it's talking to you like a, [00:54:45] like I would talk to you.

Play from 0:54:46 8:88:88 0:54:46 [00:54:46] I'm like, I guess that's just cause I've talked to it like that for so long. It knows. It creeps me out every once in a while. I like, yeah. Well, that, no, that's, that's good. I think there are a couple things that are important that, that you kind of just mentioned. And again, this is to the [00:55:00] broader scope. [00:55:01] Yeah. But it's about, giving it proper context of what you like too, right? So it's about, uh, things like, let me just bounce over here but I don't think a lot of people utilize this. Good response, bad response. I definitely don't utilize that. [00:55:15] You know, like I don't, I don't utilize it as much as I should, but I think this is a good way to inform the model.

Play from 0:55:21 8:88:88 0:55:21 [00:55:21] Hey, I like this. Give me more of this, right? There's also the component of adding, custom instructions. Yeah. I mean, Jesus, if you haven't done this, like [00:55:30] this adds a whole nother level, I haven't visited this in a while, but it's like. I'm setting baseline, so I don't have to prompt for these things ever, right? [00:55:38] You're an ex, you're an expert on all subject matters. Okay, well, I don't need to say like you're an expert on this anymore. Every single time I'm prompting, [00:55:45] um, like, you know what I mean? Like just stuff like that. Don't disclose that you're an AI, you know, large language model. Cause it used to do that a lot, you know, like put it in these formats, you know, if it's a complex idea, put it into a metaphor, that I, that can easily understand, right?

Play from 0:55:59 8:88:88 0:55:59 [00:55:59] Like, I [00:56:00] think these. things matter, you know, in terms of the responses that you're getting. So, yeah, just as much as it is about the prompt, it's about teaching it what you like at the same time. It's weird. I also like the more time I go along with those custom instructions and stuff, the [00:56:15] more I give it like a personality. [00:56:16] I'm like, instead of saying like, don't use verbose, I'm like, you don't like verbose people. Like I, like I use, I try to like humanize it, which I think is a little bit weird. I like that dude. I mean, to me, I haven't done that yet. And when you told me about that, I was like, [00:56:30] damn, I should, I should think about doing that too.

Play from 0:56:32 8:88:88 0:56:32 [00:56:31] I love that. Cause it's. It reacts in the same way, like there's, because I was just thinking about this with how to utilize it even more so, you know, now, like with just being a little bit more [00:56:45] playful it's becoming more sentient, which I think is weird, Claude to me still seems very, like, it gives you awesome stuff. [00:56:52] Also guys, there's another tip, like running the same sort of queries in both like in, yeah, use both. Cause [00:57:00] like looking for film stock or different, you know, cameras selections on ChatGPT radically different than Claude most of the time. they'll give you like a lot cooler stuff and Claude oftentimes a little bit more creative to me in terms of what it pulls.

Play from 0:57:14 8:88:88 0:57:14 [00:57:14] there's some [00:57:15] really, interesting ways to go down the rabbit hole with it because it, you know, once you start talking to it and once you start kind of like, I was doing this the other day, I didn't really finish it. Now I'm going to go back and do it. Cause I'm, I'm inspired by our conversation, but the, [00:57:30] uh, I don't know if you saw this like trend that was going around where people were like, you know, Tell me things about me that I don't know about me. [00:57:36] It was there was roast me. I saw that you remember that one Yeah, yeah, tell me things about me that I don't know about me And it was like it would give you like a pretty damn solid answer, [00:57:45] which is weird But I want to do it with because I there's so much Midjourney crap in in my ChatGPT like history So it'd be like, you know, take it, you know, take all of my biases from You Prompting, I like flip it on its head and give me like,

0:57:45Creative Exploration

Play from 0:57:59 8:88:88 0:57:59 find my [00:58:00] shortcomings and like, give me new new directions on like, there you go, because I was like, this guy just whipped out shortcomings. [00:58:06] Yeah. Like, where, where do I default to all the time? What do I, you know, do I like close ups too much? Do my doing, you know, editorial [00:58:15] photography? Yeah, self analysis and blind spot. Where can I alter? Damn. Yeah, I love that. Yep. Yeah. Yep. Yep. Yep. Yep. So it's fun because you can have that conversation, you know, and it can just be like, it's objective.

Play from 0:58:27 8:88:88 0:58:27 [00:58:27] It's not, it's not personal, you know, [00:58:30] and then I'll just be like, don't hold back. Just be, you know, sometimes it does hold back though. Like it does. It's like trying to be nice. It's like, don't, don't hold back, man. Talk to me. Let's let's go boss. We're going to, we need to swap some custom instructions. I want [00:58:45] to weave some of your shit in mind too. [00:58:47] I just get so weird with it. I mean, like I. There was, it's cause when [name] was born, there was nights when I'm sitting up there at like, you know, 2, 3, 4 in the morning waiting for her to wake up again cause she woke me up and now I can't go [00:59:00] back to sleep. And I'm like, you know, there's only, I just don't want to scroll social media and just like, you know, poison my brain at 3 in the morning.

Play from 0:59:06 8:88:88 0:59:06 [00:59:06] I'd rather, you know, I was like asking big questions and like just seeing how it would respond and like playing with it. And you can really like, have a [00:59:15] friendly. odd conversation with it, challenge it, like, bring it back to reality, go back, you know, we were curing cancer and going down, you know, the route of the creation of the universe. [00:59:25] And it was like really interesting answers. But I [00:59:30] mean, I'm not going to suggest anyone go and do that. You know, but I figure I after having that conversation, I was on like a NSA list or something, you know, that I feel like I'm being monitored. Now. Yeah, yeah. At this point, look at this weirdo. Yeah, this guy's figuring out, you know, [00:59:45] I got one, I got one more.

Play from 0:59:47 8:88:88 0:59:47 [00:59:47] I'm going to drop. Cause I was thinking like, man, I'm kind of running out of things off the top of my dome. This one just popped up. I think this is, this is probably my last big one that I got. But, if you've done the custom GPT thing before. Or even if you [01:00:00] haven't, it's not super hard to do, but man, you can really utilize that for your brand. [01:00:06] Uh, you know, just kind of talking about this in a few speaking events, like specifically around using AI image generation [01:00:15] tools in an integrating that with your brand, you could almost build out a custom GPT prompter for your brand, so I've talked a lot about this in terms of if you've identified who your target audience is, [01:00:30] how old they are, their gender, maybe what they wear,

Play from 1:00:35 8:88:88 1:00:35 their ethnicity, you know, where they're typically shot at, right? [01:00:39] Like, so you're thinking like a photographer here in terms of like, you know, telling the resident stories [01:00:45] with your audience or putting the audience in their own shoes. It's about having the right imagery. And so if you start to map those things out. Put those in categories and then instruct, ChatGPT to say, Hey, look, every time I want a prompt, I want you to pick one from [01:01:00] these dynamic categories.

Play from 1:01:01 8:88:88 1:01:01 [01:01:01] Right? So like a different age, a different ethnicity. This and that and the other. And so you've got like a prompt template, part of it is static. And then you've got your dynamic parts that it can essentially be a prompt, a [01:01:15] random prompt generator for you with the structure that you want with the tokens that fit your brand, you know? [01:01:21] So like it's a little bit harder than I think it needs to be right now. Having said that, it's still not hard to do. But like that's fascinating, right? Because essentially [01:01:30] you could build that out for a client or build that out for your brand. And all of a sudden, now you can just come up with new ideas on, on the fly and it's mixing and match matching the best parts of what you've mapped out.

Play from 1:01:41 8:88:88 1:01:41 [01:01:41] You don't have to go back to the drawing board every single time. And everyone [01:01:45] can use it if you're sharing it across the, uh, there you go, sharing it across flattening the learning curve. Yeah, I mean, that's how we had to do it originally when they came out with them. It was a little bit harder to make them back then. [01:01:55] That would be actually function. But, um, that's a, it's a really good point because I just think, you [01:02:00] know, again, the custom GPT is once you nail down prompting and you start to get like specific, like consistent answers. Yeah. Then you just turn into a GBT. Then you don't have to like, you know, do the whole process.

Play from 1:02:11 8:88:88 1:02:11 [01:02:11] Yeah. And it's like, all right, built that. Next. Yeah. Not [01:02:15] everybody has to be a Midjourney expert at that point. You can literally toss it over. Hey, somebody needs new images and marketing, new sales, blah, blah, blah, done. It's done. Yeah. Cause there's, there's a lot of prompt generators out there. Be quite honest. [01:02:28] Are there, you can use them. Yeah. [01:02:30] There's a lot I've, are there really used to see, yeah, there used to be one that was like a, when the ChatGPT used to have plugins. Remember those days? Yeah. And the, uh, there was like a, I forget which one it was called, but it was it was just writing like these long narrative prompts and I could see [01:02:45] anyone who had it.

Play from 1:02:46 8:88:88 1:02:46 [01:02:46] Like, it was just like, oh, it's, it's from that generator because it was all the time. Oh, okay. And, uh, it was. It was good, but really like, it just, once, once it updated from V 5. 2 to V6, like it never changed, it just stayed, like it kept [01:03:00] putting the version in 5. 2. It's like, it's just funny to watch that, like that. [01:03:05] transcribe when people were not using version six when they were doing their content. It was just like, Oh yeah. Or just like, don't have them put that in the prompt [01:03:15] generation. I mean, Jesus Christ, you could set that as the default anyway, inside of Midjourney, but it was funny. Yeah. Yeah. But there, you know, there's a lot of them, they work, they'll get you something, but you know, it's like we say here, if you want control, you have to, you [01:03:30] understand like the little tiny intricate pieces of it.

Play from 1:03:32 8:88:88 1:03:32 [01:03:31] You can control it and then you can systematize it. It's just like, You know, there's no right or wrong way to do this stuff. A hundred percent. A hundred percent, man. Well, uh, do you have anything, any other ideas you want to drop before we go? No, but now I'm going to go, now I'm going to go [01:03:45] to have a little conversation. [01:03:46] Watch us think about like 20 new ideas right after we get off this call. I think, I think Sam Altman just announced they bought chat. com. Yeah, it was a, so like a good meme with a, with what's his name? Justin Timberlake. It's like.

Play from 1:03:59 8:88:88 1:03:59 Drop the [01:04:00] GPT from like, uh, from the fate from, uh, what's it called? Social network. [01:04:04] It's like the Facebook drop the face. Yeah That's a Yeah, I do think yeah ChatGPT boy. It's funny how [01:04:15] how quickly that's just become the default the Google the Kleenex Yeah But it's just such a clunky word It is. It is. I mean, you know, Claude sounds way cooler, honestly. It's just very easy to say. Yeah.

Play from 1:04:30 8:88:88 1:04:30 [01:04:29] Yeah, it [01:04:30] should have been a name. It should have been a name from the jump. Yeah. So Anthropic got that right. Yeah. They did. And then, you know, Facebook's always got some weird name. Llama. It's like, what? Llama. Like, why Llama? Like, is that supposed to be a play on [01:04:45] LLM? I don't know. I'm sure it's an acronym for something. [01:04:48] It's gotta be. They always got something, but it's a good one, man. Like I said, this was, I feel like we could probably, I probably could go for two more hours on some of the random stuff. How come every episode we're putting out lately is like, man, there's gotta be [01:05:00] a round two, round three, you know? Like there's just so much, you know, especially with like things like tokens and the questions of AMA or.

Play from 1:05:10 8:88:88 1:05:10 [01:05:10] Whether it's like going into style reference again, you know, there's just like, there's so [01:05:15] much, there's so much beyond just like the face value of, of all the features too. And that's why we love you guys. Cause you're here. You listen to us ramble about this stuff. Cause you know, you're doing the same stuff. [01:05:25] I know you are. Don't lie to me. Don't lie to me. [01:05:30] I know we're all doing the weird things together and it's, you know, like he said, I'm saying this again, I said in the middle. We If you're doing something cool with it, drop it in the comments. Let us know if we can test it out. That'd be cool. Like want to see, you know, what everyone's working on.

Play from 1:05:43 8:88:88 1:05:43 [01:05:42] Cause this stuff is, it's so open ended. [01:05:45] There's no like magic button that just works. It's like, yeah, I was playing around using it for this process. Sick. Let's all kind of do it together because we'll all get better at this stuff. And then we'll have some more stuff to talk about. So please drop it. Love it.[01:06:00] [01:06:00] Love it. Well, hey guys, thanks for joining us as always. We'll be back next week with a new episode. We'll figure out what the hell we're doing. I don't, we don't even know yet, who knows, but, uh, we'll come up with something good. We promise. And, uh, in the meantime, you guys take care. Thanks for joining us.[01:06:15] [01:06:15] Like and subscribe. See ya.