Transcription
What if you can create ads with AI that are winning? What would that mean for your business? Well, I'll tell you what it would mean. It would mean that your ROAS would go up, your return on ad spend would increase, and it would mean that you might have a really profitable business. So it's incredibly important to know the skills, how to prompt ChatGPT 4.0, which is this new image model that everyone is talking about, to get the most out of it. But the problem is, how do we know how to prompt these things?
Well, I brought on a guy who's world-class at creating winning ads, winning memes, winning visual content from ChatGPT 4.0, so that I can learn how to do it and you can learn too. Enjoy this episode. This is really fun and creative, and I think you'll enjoy it.
Start time [Music] We got Jacob Posel on the pod. This guy is going completely viral with all these ads that he's creating using ChatGPT for, uh, image search. Had to bring him on the pod. Jacob, by the end of this episode, what are we going to learn?
Thank you. I'm happy to be here. Um, we're going to learn, uh, we're going to teach people how to create ads, um, ready for the ad account using ChatGPT. It's in about 2 minutes; it's that simple. And not just any ads, banging ads, right?
Yeah, yeah. I mean, these ads are sick. These ads like blow me away every single time I see them come through.
Cool. So we're going to show four ad examples and give away the prompts, and then at the end we're going to do two live cooking examples to show you how to do it. And yeah, let's just get screen sharing and and get people give people give people the sauce.
Yeah, let's do it. Let's do it. Um, so I guess something to be aware of where when it comes to these, um, to CH using ChatGPT compared to image examples people might be familiar with like Midjourney, Flux, um, Ideogram stuff like that, is because it's it's a it's a multimodal language model. It's not just a chat; it's not just an a pure um image model. The prompting is way less important, um, and the prompting you can actually be pretty straightforward, and it has a much better understanding um of intent and creativity. So the important thing I found in creating good ads are actually the inspiration images and the product images that you provide.
So my general advice when creating ads, and we can see this first one, which is um a really cool one that I created for Jones Road Beauty, um, is um providing an inspiration ad that you wanted to adjust or you wanted to use and then um providing a your own custom image. So we can read through the prompt, and you can see it's actually fairly basic. It's just, "Add this image to the top of the ad with the product image overlapping the bottom of the image and the heading and subheading below." Um, right. So we can see the original ad, which is um the Jones Road Beauty ad that I pulled from just an inspiration source, and then the image that I'm referencing, just an image from their website of a woman kind of displaying that product. And if we take a look at the actual ad itself, it followed those instructions really nicely because GPT 4 is like a really, really intelligent model, and so it understands my intentions really well. So you can see we have that woman at the top; we can have we have the overlapping products overlapping the bottom of the image, and then we have the text below. Um, so you can see like taking an inspiration ad or taking kind of a reference ad and then providing your own imagery is a really nice way that you can um create super consistent, reliable ads by just uh making edits or adjustments to them. And if you really wanted to play with conversion, you would just take, you know, a hundred characters, you know, men, female, old, young, all different colors, and see what converts for different audiences.
Right, exactly. Yeah, so I mean, um, you may have certain um key audiences or personas that you're trying to target with your ads, um, and you can just you can generate your own images because ChatGPT is amazing at creating photorealistic images of different audiences, or you can pull in your own from your own photo shoot and just um throw them in here. And when the API becomes generally available, this is going to be trivial. It's literally going to be like, "This is your inspiration or like reference ad; these are all the different audiences," just like boom, boom, boom, smash them out, and you'll have 100 ads waiting for you in like five minutes.
Yeah, and why that matters for people is because ultimately, you know, the game is how do you create ads that convert, and it's really a game of experimentation. So what this does is allows us to experiment at scale, and you don't know if it's like this woman looking to the left or looking to the right or with blonde hair or red hair that's going to be the difference between, you know, a two, you know, 1.6 ROAS and a 2.8 ROAS.
Yeah, yeah. Marketers have this saying nowadays that the ad is the targeting, and this is just a way that you can actually do it at scale.
Cool. I love that. Can we um I mean, this is a simple example; like I get it. Um, let's see something else.
Yeah, let's do it. Let's go to this one, which is a little bit different with no reference images. So um, this ad is an absolute killer. Can I ask can I ask you something before you get into it? When do you know to do a reference image versus not a reference image?
So okay, a super important thing to understand about these models is that they have a really good understanding of um, well-known concepts, ideas, brands. So it knows Nike really well; it knows Adidas really well, um, even like the classic online e-com brands like it knows Ridge Wallets really well. And so in that sense, you can prompt it, and you can just say like, "A Ridge Wallet," you know, um, and it will understand what you're talking about. But if it's your own product that is relatively new, relatively small, um, and you don't have a lot of online presence, and I'm talking like a lot of online presence, like millions of dollars on ad spend a month kind of um presence, then you should always provide a reference image. Um, but for like Adidas, everyone knows Adidas, and so it's going to do this; it knows it really well.
Cool. I didn't know that actually. So that's helpful.
Yeah, yeah. It it has a really good understanding of certain things, and others it just doesn't know. All right, let's let's rip through the prompt. So this prompt um is one from um Sora. Uh, Sora is has like their own explore page, and I found this one, and I thought it's really, really incredible, um, but it and it has a lot of really good lessons in here. Um, so you can see using the "ultra realistic" or "photorealistic" keyword is really important for creating photorealistic images; otherwise, it can sometimes default to something that looks a little bit cartoonish um or weird colors. But "photo realism," it is strong, but using that keyword is important. Um, and then you can just see it's a really strong prose description of the image: "A female runner is captured in mid-stride running laterally in the foreground wearing her performance gear. Um, behind stands the um the Egyptian desert with the pyramids." Um, so this one is fairly long, but it's a really detailed prose description of your desired image, um, but you can see it has some really nice other elements like um font type, Exo 2 Bold Italic, um, stuff like that. For similar concepts for well-known fonts, it is going to have a decent understanding and representation of what those fonts mean because it's seen enough of that content on the internet. So people can mess around and play around with that. Similar concepts are like in order to making adjustments is like styles, um, like um color styles and also um aspect ratios. You can make adjustments in the prompt to aspect ratio similarly. So um, here's the ad. You can see it's very creative; you can see the woman running um with the pyramids in the background in the Adidas shape, um, something just really creative and really nice, um, and that was like literally impossible to do before.
Sorry, did you say that you use Sora for the prompt? Is that what you said?
Um, so I highly recommend people check out Sora. Um, Sora, it uses the same underlying model, but they have an explore page where you can see people; it's a really good way to understand how people are prompting and how the best prompts work and to get um and for ideation. Um, if you click on any image in their explore page, it shows you the prompt. So can we just we we check that out real quick and do a little detour?
Yeah, of course. This is a piece of alpha that people do not know about right now. I don't see anyone talking about it. Another really cool thing is I think Sora is better um at it's has less strict policy guidelines, and um it's better at photo photo realism. I don't actually understand why that's the case, considering they probably use the same underlying models, but um I think that it's less strict and um yeah. So but if you can see, so an image came through, for example, um of of this guy, uh, kind of Post Malone energy.
Yeah, yeah. Like, but hyperphotorealistic, like that is that could be a photograph, you know, um, and we can click on it, and you can see the prompt: "If Weird Al and Post Malone had a baby." Yeah, um, but you can click on it; you can see it, and you can also adjust it or edit it. Um, so I could come in, and I could remix it from here. I could adjust the prompt myself, you know, so I could like literally just start rewriting things, um, or I can change the aspect ratio, um, or I could upload my own reference images. Um, so yeah, very um very, very cool. You can see here I could like upload a reference image to make adjustments, but it's just a really, really great place for people to go through and just scroll through and understand, look for prompts that may work for them. Some of them are ads that um you'll be able to see, and you can make your own clever adjustments that way.
Yeah, I think what's really cool about this is this gets your creative faucet going. So um, I think the game right now is is is looking for an aesthetic that isn't being used in your space. Um, so if you can figure out how to do that and just sort of explore this and be like, "Oh wow, this would actually really this like picture of um, you know, the Grand Theft Auto thing would be really cool as a meme in whatever space that I'm in," boom, all of a sudden like that's gonna get you a bunch of likes, going to get you a bunch of followers, going to give you affinity. So that's what the game is about, and and I love this because I'm always struggling; I'm actually struggling with the prompts to be honest, like I don't know; I know that, you know, sometimes I'll throw in a word like "hyperrealistic," like it takes me a while basically to get to that word, "hyperrealistic," um, and I'm just like wasting time. So I think what's cool about this is it shortens the time to value and also gives you ideas that you wouldn't otherwise have have had.
Yeah, yeah, yeah. Absolutely. I think that like the the reference images is a really good way to shortcut like the prompting um issues. But if you need help with prompting, I mean, um, it's it's really important to be very clear, very direct in what you want. Um, and another piece of um another piece of information is if you struggle with the initial prompt, it's important not um to carry on in the same conversation and try and adjust your image in the same conversation because you generally get the recycle of worse and worse um rather than it actually improving because it's using the generated image as its reference for the next one, um, which can have its own inherent flaws. So I always tell people if you want to adjust a prompt, start a new chat, um, and make the adjustments to your prompt in that new chat, and the images are going to be consistent enough that it's going to be similar to the one that you first got with the improvements in your new prompt. So that's something else that um is is important when trying to prompt and create these things.
Should we move on to the next example?
Yeah, let's do it. This is another example of product image um plus reference image. Um, so you can see I found this original reference image for Rise. Um, I then took the Athletic Greens product imagery, and I just gave it some really clear prompts. So I said, "Recreate this ad for AG1, adjust the background to better fit Athletic Greens style, change the price you can see over here to $110, slash it out to $85 now," and then I said, "Generate the image." The reason why I said that is sometimes it will try and generate text which represents the ad. If you just say "generate the image," you're going to consistently get um image results. And I mean, if we look at this ad, it's sick; it like it really got it well. Um, it looks amazing; it got these like icons over here; it even got this like cool gradient color in the background, which is really incredible. The background, as you can see, placed the images which the products which look incredible. Um, it got the pricing right, and the copy is pretty nice, and like it fits really well; it's scented and everything. Um, so I mean, this is probably a runnable ad, and it it it was very, very easy to make.
Yeah, and it's, you know, a couple things going through my mind: one is like it somehow feels dirty though, you know, it's like in the sense that and and I'm curious your perspective on this, like that other company, Rise, spent all this time and money, maybe even tens of thousands of dollars figuring out this ad, an ad copy, and then in like a few seconds we're I mean, we're kind of we're remixing it into, you know, for AG1. So I'm curious like how do you is it dirty, or am I just too clean here?
I mean, to be honest with you, the way that most of these um companies work is they are finding inspiration ads; there's like 50 companies that do this; they provide you inspiration ads, or they're getting templates which are recreated ads in Figma files or Canva files. There's companies that basically just do this um at scale, and this is how people are making ads right now. So I don't I don't think this is dirty because I think this is was already happening. Um, it was just happening physically; people were looking for inspiration; they were um they were using templates. Now um ChatGPT just lets just does it for you. So I mean, from what I've seen, this was already happening; this is already happening. Yeah, this just makes it a bit more accessible to anyone, but it was already happening.
All right, let's let's move to uh the fourth example.
Yeah, let's do it. Last one um is I wanted to show was changing the concept of an existing ad. So um I found this Ridge ad, which was um which is kind of a Black Friday ad with their gunmetal wallet, um, and I swapped it out for this green um Ridge wallet, and I said, "Swap it out and change it to a Christmas-style ad," and I mean, there we go; it's definitely a Christmas-style ad for this wallet, um, and yeah, you can see again, the prompt is because the reference images really, really do a great job of um of replacing the need for super strict prompting. Um, it's like a picture's worth a thousand words; that's literally the case here. Um, so very simple prompt, and we can take a look at the ad; it's pretty it's pretty great. What do you think?
I think it yeah, I like it. It just gets me thinking that there's like there's the big T-pool, you know, major holidays, the easers, the Christmas, the New Year's, Valentine's Days, and there's probably a bunch of like little funny ones, you know, like National Hot Dog Day and stuff like that, which you're there's probably alpha in and focusing on that. There's probably also alpha and targeting people on their birthday. It's like, "Happy Birthday 30% off by Ridge Wallet." Oh my God, that is that's that's a great idea. So I think I think this is this is what excites me about all this, right, is is I mean, is the fact that the targeting is going to get a lot more personal; the it's going to be a lot more funny; it's going to be ads are going to be a lot more entertaining, and it's it's really up to, you know, the idea people to come up come up with these ideas. I'd love to show you this one that I put together um which which relates to that, which is um I I took an ad um I wanted to do one for Rippling. Did you see the the Rippling deal controversy with the spy? Um, so we'll wait for this one to load, but you companies now have the ability to take advantage of um short like news cycle moments like they never could have before because it was expensive, prohibitively expensive and slow. So um I wanted to make an old Wall Street Journal, Star Hill newspaper ad, um, and it's Rippling, "So good our competitors can't even resist sneaking a peak," which I thought was pretty funny. Um, but yeah, it's like it's a different style to what they'd be able to do before; it's got really clever, funny copy, and it takes advantage of a moment, you know, and this could you could have AI just like ready to do this, like watching Twitter, watching um some other news feed that you want and just like it's just on it, and these could be produced on mass. And where you know you mentioned I think it was with the um well actually with a lot of these you're getting these ads, so where are you getting these ads? You going on Meta, Meta's ad library to get these ads? Like, where could people basically get ad inspiration?
Yeah, the one that I use um is CreativeOS; they have their own inspiration and templating tool. Um, some other good examples are for Playay. Um, I think the recent um Icon.me is one that recently launched that has all of these um it's basically becoming a staple of any ad building tool. Um, so they they basically read in the Meta ad library; they tag them, and they make it easy to search over them. So CreativeOS, for Playay, those are some good ones to check out.
Cool. Um, those are our four examples. Um, do we want to do just live cooking one or two just to just so people can get in in your head as you know you're coming up with ideas and what you're thinking about as you're doing it?
Yeah, absolutely. Let's do it. Do you have any um any companies you want to you want to use or target? I mean, we can use my you know my own company I guess or so um like I'd love to create an ad for one of my own companies.
Okay. Which one?
Um, how about LCA, which is our design agency for the AI age? Um, so we work with like Dropbox and Salesforce and figuring out what did their AI interfaces look like and help them figure that out. Um, we've never done ads before, so it would be cool to try it.
Let's do it. What's the domain?
LateCheckout.agency. It's like literally a landing page. I'll I'll talk people through exactly how I'm thinking about this um and like the important elements of an ad. So um the important elements are this logo up here and specifically kind of your typography and this use of bolding and italics, you know, like that's really important to capture, um, as well as the coloring and then also some of these um these these like maybe these brands that you work with. Like I really like this aesthetic, so I probably begin by if this is an image I'll copy it. Okay, it's not, so I'll just I'll just take a screenshot of it; it's fine. Um, like that I just took a screenshot, and then I'm taking a screenshot of this too, um, and then maybe kind of this down here as well. Um, and can you see that I'm taking screenshots by the way?
No.
Okay, I'm taking um I've taken screenshots of of this content over here, the stuff over here; I've taken a screenshot of this content and of the logo too. Um, so next thing I might look at um I would think like, okay, do you have any inspiration sources that you would want to use um for for the ad, um, or do we want to kind of just like um do we want to kind of just like YOLO it?
I think like the newspaper style theme that I put together could work really well for this because you've got that like paper theme, um, but you tell me what you think.
I mean, I'm down to try that. Um, okay, from a brand perspective, like LCA works with only a dozen companies per year, and it's if you look like the whole vibe is like it's almost like we're a SWAT team, like and for your executive team, right? So it's like pretty working with SVPs of Salesforce, CMO of Slack, so it's very it's almost like we're selling a luxury product. So when I think of like brand inspo, I'm like this is more like a Patek Philippe, you know what I mean? And what I don't like about the newspaper is like the newspaper is like, you know, when you're thinking of design for for AI, you're not thinking about a newspaper, right? This is like futuristic.
Yeah, yeah. No, absolutely. Okay, let's do you want that like like that vintage style or do you want like a futuristic style? 'Cause I'm thinking like one of those old vintage Rolex um yeah ads could look really great here.
Yeah, that would be really cool. Okay, let's um if I if I go to another tab, I'm so typing "vintage Rolex ad." Okay, we can see if there any here that we like, otherwise um tell me if there's any that are like kind of um that that you particularly like, or we could ask or we could just try prompting to do one.
Oh, go down, scroll down. The guy pointing, this one.
Yeah. Okay, let's try. I mean, why? Like, I just like the I don't know, the color stand out, and it's kind of like it's like, "Hey, do you need AI?" You know, "You need AI interfaces." He's a bit sad, and and from a conversion standpoint, like you you often don't want like sad, angry people; you want like happy people.
Mhm. Okay, sick. Let's do it. Um, so we're going to go to ChatGPT; it's just loading; there we go, and we're going to make sure that we have GPT 4.0, and now I'm just going to like bring in those images. So I've got this one; I've got this one, and I'll show them again, um, and then this one. So we've got um kind of the who you work with, um, we've got the some other kind of copy and the logo, um, and then we've got the inspiration ad. So couple things to be aware of that may cause this to not work well: we do have to specifically prompt it to write English; otherwise, it might get confused there, um, and the text is a little bit kind of like low-res over here. So let's let's see if we can if we can try and make sure that um that comes out well. So do we need to say like this is a print ad, and we're you know we don't want a print ad; it's just inspo; you want to create something digitally?
Yeah, yeah.
Um, let's begin prompting. So I'll tell you how I'm thinking about it of it. So we're going to start with the initial instruction instructions: "Generate a digital ad for LCA, um, LateCheckout.agency. Um, it it probably doesn't know who you are; I'm just I'm just going to assume it doesn't. So LCA is a SWAT team slash um how else would you would you describe team? CL design agency for, you know, the most interesting brands on the planet, helping them with AI interfaces or helping them with AI building AI-first products. Um, it okay, it only works with the best companies." So because I'm telling at this it may cop it may kind of focus on this in the copy; we'll see, but that's kind of what my intuition is telling me. Um, "I have provided some screenshots from their website for you to use as um inspiration in the ad as well as a reference vintage Rolex ad um as inspiration. You can use this um use this to capture the luxury um field of the company. So um generate generate a digital ad for LCA, LateCheckout.agency. LCA is a SWAT team design agency only for the most interesting brands on the planet, helping them with building AI-first tools. It only works with the best companies. I have provided some screenshots from their website for you to use as inspiration in the ad as well as a reference vintage Rolex ad as inspiration. Use this to capture the luxury vibe of the company." Um, so we've got kind of main instructions and context and some more specific instructions. We are a little bit uh vague here because we haven't given it strict instructions, but that's also because I don't actually know exactly what I want to get out of this, and so I'm hoping for it to leverage some of its own inbuilt creativity. Um, so let's um just say "generate the image," and and by the way, before you hit go, could you say "generate three different versions of this"? Is that crazy?
Right now, um, I think ChatGPT just does one at a time, and I think that's on purpose because their GPUs are kind of going crazy. Um, Sora, you can control the number that you generate; I think you can generate up to four. So Sora is like basically the exact same thing, and maybe we could show people afterwards, but I'm just curious, just curious.
Yeah, yeah. Um, that's 100% coming. Cool. Okay. Any last adjustments you want me to make before we go for it?
No, I trust you, and Jacob, I trust.
Okay, let's see. Um, so this is what I would generally say is like a higher-risk ad, you know, um, when I prepared some of the other ones, like I knew it knew what Ridge was; I knew it knew what AG1 was, um, and I gave it really strict instructions, um, but I'm really curious and really interested to see how this does. Um, it's kind of like relatively vague, and we're kind of relying on its creativity um in a relatively less well-known kind of company or product. So um I'm actually really excited and curious to see how this does, and I think it will be really relevant to people if it can do well. Um, so it starts off fully blurred, and it's going to slowly generate from top to bottom; that's a good sign, and we can just kind of wait and see how it goes. And how often are you oneshotting these ads, you know, or is it more of it takes three, four, five, seven back and forth to get to something that really works?
Depends. Depends. For something like this, it's usually like two or three times, and I think that's fair to to kind of assume it's not going to oneshot my intention or my taste every single time, um, but for some of the ones that I demoed, that was first try because again, it's it's low risk because I gave it strict inspiration, um, I gave it um I gave it a well-known product that I know that it understands, and it's an ad that fits really well um for the kind of product that I gave it. But this, we're kind of mashing up together a lot of different themes and a lot of different styles.
Dude, I'm loving this already. Generate.
Yeah, it looks sick. It kind of reminds me of uh some of the Perplexity AI ads; I don't know if you've seen those.
Yes, yes.
Oh, that's that's good copy, or at least I like it: "Leaders at billion-dollar brands choose LCA to define their future." I like that. Yeah, so we got the man pointing with the fire in the background; he looks like he's at a golf…
Course, or something. Yeah, this is the craziest game of golf of his life. Um, I hope the logo looks pretty good, and it got that typography nice. Oh, it's generating the watch. Watch the watch. Okay, so that's something. Now we know um that we don't want the watch there; that we want to replace it with some other product. Um, it it it adjusted the copy for us, but it changed the brand of the watch, which is funny. I might have thought that this is a this is a watch brand, but okay. This is like important; this is something to understand. This is what we know. Um, and um, okay, sometimes it gets sometimes it gets caught like this, so but it tells us that the image was created, so I'll just refresh the page, and it should generate the whole thing. The LCA Rolex? Hilarious. That is hilarious. Um, okay, this is either a combination of my internet being weird or um, OpenAI is struggling, but this will come through. I saw Sam was like, "Our our GPUs are melting," quote unquote. Yeah, yeah, literally. I think like this this moment, okay, there we go. Um, this moment for them was probably as big as ChatGPT itself. Yeah, but I think that looks that looks really good to me. Um, it it like kind of got the right vibe that we wanted. It got the copyright; it turned it into a more digital style ad with like the CTA over there.
Um, what I would do now, um, and this is a really important principle to understand, is I would take this, um, and I would tell it—I would do this in a brand-new chat—Do you want do you want me to go through the next stage here? Um, let's do it. Um, so I would literally—this is going to be a little bit of me stumbling around—let me just make sure all of these images are saved. Um, so instead of you just replying in the same thread, we're going to open up a new thread. Yeah, and yeah, it's not like entirely entirely the most important thing in the world, um, because we haven't provided it a strict reference product to go off of, but I personally have found that it produces um superior results because it's going to be using that previous image as its reference, and that image is still AI generated. So the text is still a little bit worse than P than your um text here. The logo is still a little bit worse because it can never be more perfect than the reference; like that's kind of inherently impossible, and so it's always going to get a little bit worse over time. Um, so um, what do we want to replace the Rolex with? Do you have any ideas? I mean, we can—a couple things come to mind—like one is we can show some of the some of the design work that we've done, but I feel like that might be a little too like I don't know. What do you think? Is that a little too obvious? Should it be a bit more abstract? Um, I don't—I think it would be cool to maybe show the companies who you've worked with. That's like maybe let's do that addition, um, rather than generating a watch, show some of the companies LCA has worked with, provided in the reference images. Okay, so I've created a new thread, and I'm giving further instruction. So, um, you can see we got the original stuff. Um, rather than generating a watch, show some of the companies LCA has worked with, provided in the reference images. I'm just like making sure that it knows um making sure that it knows where it's coming from, so it doesn't think to hallucinate or make that up. Um, and I think that's it. So any last additions you would want to make before we go for this? Let's let it roll. All right, let's do it. So how different do you think this is going to be like? Do you—are we going to still get the same dude up top basically, or like what are you thinking? I would assume so. So there's like this concept of temperature when it comes to um LLMs, where temperature is the amount of randomness that comes out of each input given the same each output given the same input. Um, and I think the temperature on these is quite low, so um generally if you give the same prompt, you can expect a similar response. Um, so for example, if you generate people, um, and you use kind of a similar a similar prompt, the people will come out looking quite similar. So I I I anticipate that it will be quite that it won't be very different, but the image will be um the the like Rolex should not be there anymore. The watch should not be there. Cool. And then from like a dimension standpoint, right, like I want to be able to take this and actually like run it on Meta, run it on different platforms. Do you have to specify dimensions? Um, it usually defaults to 9 by 6, but you can specify like 1 by 1, 9 by 16, um, different aspect ratios. Cool. And while this is loading, I'm peppering you with questions. Yeah, the um like have you played around with taking some of these images and turning them into video content? Because a lot of these ad platforms are running video, right? I I've seen people take my ads and turn them into um into video, like animate them at least. Um, a lot of people will do that for kind of like um for um like DPA ads, like the display um ads, but um I haven't turned them into super strict um into like pure pure video, but I've seen animation of them. But um I would recommend testing out like Runway and Sora and stuff like that and being creative, though.
Okay, we can see it is a little bit different. We now have LCA. Is he wearing a Rolex? It looks like it. Yeah, his um his he's no longer pointing; he's now showing the watch, which I don't know if I'm fully against at all. Mhm. You know what I mean? Like it's a bit it's less it's less jarring; it's more of like, you know, fist bump. Yeah, yeah. Okay, same thing; I got caught. Um, so we'll just refresh, and it will come. But boom. Yeah, it's pretty it's pretty dope. It's pretty cool. It's got the it's got the logo right. Design f for the AI age with the um subheading and all the different companies you work with. Um, fortunately, you guys work with well-known companies, so we got their logos right, which is um a little hack for you guys, actually. There's one company on this list we don't work with that they included, which is Disney. Really? Oh, interesting. Not that I know of. I don't think we work with Disney. Um, okay, not that I know of, but uh overall this ad looks it looks really really dope. Yeah, it's a cool ad, and like it's not it's probably not something that you would have put together, and it's only been like 10 minutes of us testing it out, and we put it together. Um, if we wanted to get more um if we wanted to get it more guided towards something that you like, um I can talk through how we might prompt that. So like we may—if we do want him pointing, we can specify the man is pointing. Um, if we did want to be stricter with the logos, we would say include all of these logos, and we would specify them by name. Um, if we wanted like the previous one, if we wanted it on the side, the text on the side with the image on the right, then we could have specified that. So now we have kind of the idea of what's coming out, what might come out, and now we can specifically prompt and specify what all of the different um the different ways that we want to adjust this because our prompting was kind of like relying on its creativity, and it did a really good job there, but now we have the kind of the creative juices flowing to be able to guide this towards exactly what we want. I love it, man. This is has been fun. I appreciate you uh dropping the sauce, giving the frameworks, giving the principles, giving us uh an inside look into your head. I appreciate the free ad that you generated for us. Share this with my team. Um, Jacob Posel, appreciate you. We include links to follow Jacob in the show notes. Um, appreciate you, and uh if people want more of the live cooking around ads, how to create them, um please let me know in the comment section, and we can do more episodes on that. I personally believe that this is an incredible time to create memes, uh ads, uh creative like this and stand out. So um now's the time. Appreciate you. Thank you. Thank you. It was a lot of fun. All right, my man. See you.