Mastering Midjourney: How to create consistent, beautiful brand imagery without complex prompts | Jamey Gannon

It all comes down to having a very tight and manicured process, which thankfully I have spent my 10 gajillion hours in MidJourney and NanaBanana and everything to figure out exactly what that is so you're not pulling your hair out prompting all day. One of the things I like about the mood board is it's a visual language to explain to MidJourney what you're trying to do. The picture is worth a thousand words. Like literally a picture to an LLM is worth a thousand words. Mentioning like Vogue or high fashion or even like a different artist's name is a great way to tell the model a ton of stuff without actually having to tell it a ton of stuff. In the past brand and creative directors or agencies would give you these photos and be like cool, call us and re-up when you want more photos. What I love is that you're like, look, you're going to value me for all this upfront work that I'm going to do to define the space, give you these codes, really give you reference images. And then now you can go do this for yourself. It's just like a very different model of providing service. And I think it creates a really positive collaboration between the client and the creative director. Welcome to How I AI. I'm Clara Vo, product leader and AI obsessive here on a mission to help you build better with these new tools. Today, we have an aesthetic episode with Jamie Gannon, who is an AI creative director and is going to show us how to create consistent, beautiful and unique brand assets using MidJourney, Nano Banana, Flora, and more. This is a workflow we haven't seen yet and goes into incredible depth on how to create awesome brand assets that you can use to uplevel all of your designs. Let's get to it. As an AI founder, you're used to sprinting towards product market fit, your next round, or that first enterprise contract. But speed isn't enough for AI startups. Buyers expect security, compliance, and transparency from day one. That's why serious AI startups use Vanta. With deep integrations and automated workflows built for fast-moving AI teams, Vanta gets you audit ready fast and keeps you secure with continuous monitoring as your models, infra, and customers evolve. AI innovators like LangChain, Rider, and Cursor scaled faster and closed bigger deals by getting security right early with Vanta. Listeners can claim a special offer of $1,000 off Vanta at vanta.com/howi.ai Jamie, thanks for joining How I AI. I am having you on the show for a very selfish reason, which is I think I'm the only pink AI brand in all of SaaS. And when I saw your work, I was like, oh my God, I need this lady to teach me how to create brand imagery that is beautiful and fun and girly. And whatever magic he has, I need. So talk to me about how we can get amazing images like what you're showing us right now. I think consistently is the most important part because I can get a one-off image that's this great, but I can't get this brand portfolio. So you got to tell me your sorcery. Yeah, I mean, it all comes down to like, unfortunately having a very tight and manicured process, which thankfully I have spent my 10 gajillion hours in MidJourney and Anna Banana and everything to figure out exactly what that is. So you're not pulling your hair out, prompting all day. And yeah, I would love to show you. Great. So where do we, where do we start when you teach people how to do this? You know, where, what's, what's step one? So first thing that I always do is I'm going to start in either Pinterest or Cosmos, and I'm going to create a mood board that is the general vibe of what I want. So for this exercise, I wanted to have like a very pink and cute, but still kind of like not super girly, very internet kind of coded aesthetic. I like, especially with AI doing like juxtaposition. I think that's really fun. So we have like an orange with piercing. We have like a fluorescent fruit, dog on a computer, things where they shouldn't be. We have like a grungy unicorn. So that this is like a really cool aesthetic to start with. And I typically start, there's two ways that I usually start. I will either go in with a mood board in MidJourney and you can basically just copy and paste your images in, or you can start by adding them as Srefs, as you can see here. So Srefs basically are style references and they kind of do exactly what they sound like. It just tells MidJourney to take the overall style and coloring and camera treatment and vibe, if you will. And it like tells it to apply that. Now, one of the first prompts that I tried to create this aesthetic is I use that mood board that you saw that I made. And in this part of my process and the create part of my process, I'm just trying to get information. I'm trying to figure out like, what are the images telling AI? What is my prompting telling AI? What's the mood board telling AI? And I just want to generate very fast. So I'm not very precious when I'm doing these prompts. Like right here, I just have beautiful female model. I have astronaut. You can see I'm using that mood board here. But if we remember like the original mood board and those final images that you guys have a sneak peek of, we can see that we're like very far off from where we want to be. And I think where a lot of people that are starting to use AI get kind of tripped up is like, if you've ever just like raw dogged generated something in like MidJourney or ChatGPT, like this might be like great to you. And like some of these images stand alone are really cool to me. But if we're working for like a client or we're trying to be consistent with the brand style, we need to be like really, really honest with ourselves on like, does this actually look like that vibe? And truthfully, it does not. So I think that there's going to be a better approach to get things going. One thing I want to call out, if you could go back, is I think one of the things people lack when they're working with more of these truly creative generative AI tools is they lack language. And so one of the things I like about the mood board is it's a visual language to explain to MidJourney what you're trying to do. What I like about the style references, which we've done a couple episodes that have referenced style references in terms of MidJourney. And we've done one with Zach, the creative and design lead at Gamma. We've done one where we were looking at style refs for more photography styles. So these are just alternative languages to tell MidJourney or another tool something specific about a visual aesthetic, which I feel like a lot of people just aren't trained. If you're not trained as a designer, if you're not trained as a photographer, you don't have it. And so I love this. A picture is worth a thousand words. Like literally a picture to an LLM is worth a thousand words. The other thing that I want to do, if you go to your mood board versus what was generated, one of the things I want to call out about this, where you're comparing the mood board to the images, is somebody who has been a designer and has been around photographers. I bet you can see this and say, okay, like the saturation and the contrast on these photos is not as high as they are in the generated images. There's this like washed out vibe on some of the photography on the generated images. And so one of the tricks that I wonder if people might think about is you could actually upload this to like a, like a ChatGPT or Claude, and you could say, explain to me why the photos don't match the mood board. And so you know, you are probably an expert at this and have language to figure out why it doesn't match. But if for folks that are trying to teach themselves language, that's just one trick I think is really useful is throw this in and say, hey, ChatGPT, explain to me why these top four images aren't in the same style as the bottom. And that can sort of give you a seed, seed to start. I try and avoid prompting at all costs in my process. But like for this example, we're going to go like super, super simple with like using Srefs and mood boards. But yeah, if I'm in one of these, like I'm doing some insane editorial work for clients that needs to be consistent across like a hundred images and like foolproof to like a consumer eye, that's when like getting really into the details, especially with models like Nana Banana can be super helpful. Great. Okay, so this doesn't match. What do we do? Basically what I kind of glean from this is like the mood board is not doing its job. It's not communicating the vibe properly. This is something that happens a lot with MidJourney mood boards. There's not a ton of like documentation from MidJourney on exactly how it works, but as creatives, you can tell, like the more kind of consistent a mood board is, let's say it's like five images of like fuzzy 3D cats, you're more likely to get an image of like a fuzzy, whatever you prompt. When we're doing more generalized vibe stuff like this, MidJourney can tend to average things out with the mood board. And I find that using Srefs as the mood board instead essentially can give much better results. So this was sort of the next step in my process. I wanted to try the Srefs and see if that made it better. And you could tell we're definitely getting better contrast. We're getting a little bit more kind of aesthetic and edgy, a little bit more of that like 2025 aesthetic that we want, but it's pulling really, really green for me. So what I for some of your earlier ones. That one's the one that stands out to me. I mean, it's kind of, it's new. I'm liking where this is going now. Hold on and side note, it's just you and me and little astronauts in the middle. Yeah. So you are combining some of your style refs and this like personalization model on top of each other. You're getting way closer to what you want in terms of iteration. So how do we take this to the next step? Yeah, so I want to show you guys some like of the other mid-journey techniques and ways again to like start prompting and kind of like upping the ante on like what we're making. So as you can see here, these prompts are no longer just like woman or astronaut. We're actually starting to give some like aesthetic thought. So one prompt, I actually found this on like the explore page a couple weeks ago, but it's day's editorial photo shoot. So Day's like this really hip magazine if you're not familiar. So doing stuff like that, like mentioning like Vogue or high fashion or even like a different artist name is again a great way kind of in the same line of like a picture's worth a thousand words to tell the model a ton of stuff without actually having to tell a ton of stuff. So with like a day's editorial is like a really famous like ASAP Rocky cover. It's like super gritty, high contrast, all these words that like are really hard to remember all of them and the aperture and stuff. I barely know what aperture means. So sometimes when I just want to try changing up the vibe or make things more realistic, make them have like a more 90s aesthetic, I'll just paste in like a camera that mimics that. So the Sony RX100, I think I probably generated this with like ChatGPT. I'm assuming that this is a 90s digital camera. And this is kind of like what I would consider the final image. So we know maybe the lights need to be fixed, but we know it's in New York. We can see it's in like a high rise. We have that couch. It's a deer. Oh, another thing too to know what the prompt is like saying luxury New York City apartment instead of saying on the fourth floor of a high rise in a new post-war New York City apartment, you just say luxury because like everybody knows and the LLM knows or nylon, the AI knows luxury is going to be like a big metal thing. Super high up and rich people live on the top. So that one word luxury just gives us everything that we need to know about the scene, which is very fun. What is great about this particular prompt is it reminds me of what Ravi in an earlier episode did talking about mid-journey for generating images again. And he's like, you need the subject and you need the setting and you need the style. And so you have the subject, which is the deer. You have the setting, which is this luxury apartment at night and you have the style. And he did a very similar thing, which is like cameras are cheat codes for styling. And again, what I appreciate about what you're doing and we love to hear on how AI is everybody just wants to be lazier with their prompting. No one wants, I mean, if you go to the explore page on mid-journey and you look through people's prompts, like people need a job, man. Like it's just too long. And so I love the idea that you're just trying to find shortcuts for yourself to make these shorter and shorter and shorter, but still get the same quality because you're generating a lot, a lot of images. Oh yeah. Oh yeah. Like sometimes like thousands a day depending on what I'm doing. If I'm lucky, I nail things really fast. Like, like this aesthetic I happened to get like very quickly. Whole point of it being 30 minutes. But yeah, if you're trying to do something that needs like a lot of different aspects, like I was doing a stock photo project recently. I'm doing like nature stuff. I'm doing people stuff. I'm doing skin. So like in this case, like I can't really have just like one mood board because there's a lot of different things that we need to do in terms of treatment. So like being able to get like super like, what's the prompt do I have here? Like CMYK highlights for me has been a big one that I've been using deep blacks, high contrast. And then we have just a really powerful Sref, which actually came from a previous mid journey generation I did that day. And then we get all this like cool stuff. I think, yeah, the prompt's experiencing music. So we're starting to see like ears and like, you know, maybe body sensations. That's another tip too is like when you're in especially the create phase, maybe even iterating phase, they kind of blend together. Sometimes I'll literally, if I'm doing like a startup deck or something like a VC deck, sometimes I'll literally just like paste in the sentence and just like see what it gives me. Like the full sentence, like this is our business model. Just to get me thinking, just get the model thinking. Same thing. Like I'll just do like financial markets, really vague stuff. Mid journey is like very poetic. I would say you can like write little poems to it and it'll actually give you some, it'll, it'll do what you want it to do. Like this is experiencing music to me. You know what I mean? That you feel like the vibrations through your body versus, you know, if you were to describe this image, it's going to sound like a whole bunch of nonsense. Even worse, even worse if you do it in like a Gemini or something and ask it to give you, you're just never going to be able to get this, which is why I love MidJourney so much because it just feels like an extension of your, of yourself basically. Every time I see MidJourney, I have to tell people, you know, now I'm this host of this AI podcast. I get to see a lot of tools and I do a lot of things every day. And I MidJourney was the first tool that just switched my mind about what was possible with AI. It really is an inspirational tool. It's fun. It's accessible to not just people like you that are building a business off of this, but my kids love MidJourney. It's such a creative space. And I feel like it's one of the more, you know, dare I say it, like soulful AI experiences. And so I love the idea of just getting out of like the tactical and practical prompting of like, I want this thing and this camera and blah, blah, blah. And just playing in a space with it to see if it can inspire, inspire things for you. This episode is brought to you by Lovable. If you've ever had an idea for an app but didn't know where to start, Lovable is for you. Lovable lets you build working apps and websites by simply chatting with AI. Then you can customize it, add automations, and deploy it to a live domain. It's perfect for marketers spinning up tools, product managers prototyping new ideas, or founders launching their next business. Unlike no-code tools, Lovable isn't about static pages. It builds full apps with real functionality. And it's fast. What used to take weeks, months, or even years, you can now do over the weekend. So if you've been sitting on an idea, now's the time to bring it to life. Get started for free at lovable.dev. That's lovable.dev. Okay, so just to recap really quickly on our mid-journey journey, we have done mood boards. We have used those mood boards to kind of create some and get a sense of what's working well with the mood boards or not. We've pulled in those mood boards via style references. We've also pulled in specific images as image references. You've shown us how to go from like the most generic astronaut prompt to slightly more specific, but still pretty lazy deer New York City luxury prompt. Now you're starting to get stuff that you want. How do you kind of like package this up and scale it out? I'll basically just keep going with this same SRF set and just continue to generate images across, you know, the subject matter. So for like this one, you know, I'm thinking about AI bubbles. I'm thinking about talking about technology. I'm thinking about talking about culture. So I literally just prompted AI bubble using the same prompt, but as you can see, we're getting the dreaded like five finger thing. So in this case, I'll hit very subtle or very strong and it'll help me like come up with a couple more generations of that aesthetic. So that's one technique I'll use kind of like fight those. I will also go and I'll like steal prompts or get inspiration for prompts from the explore page, especially when I'm generating like really large kind of like stock photo sets of images. So like here's like an edgy man. I don't have any edgy man photos yet. I thought the basketballs could be a cool motif and the aesthetic. So then I kind of like pick apart those prompts and I use them here. And then eventually I get to kind of like what we saw at the beginning, all these images that I'm very, very happy with and I'll move on to sort of reinforcing my prompts if need be and editing images. So one thing that I always try, especially once I have like the exact outputs that I want already, I'll go ahead and I'll make another mood board again. So you can literally just add from your gallery. You can just click. So I selected, I think this is about like 30 images that I liked and then what you can do is you can use this again. So what should we prompt? Turtle in the sea. I was thinking turtle. This is weird. You're thinking turtle. We're floating through mid journey together. I was like, maybe she'll do like a turtle. Yeah. Very funny. Just in case. Yeah, we'll see what this gives us, using the SREFs. Hard to see we're getting a sexy lady naturally. That is one of the downsides of MidJourney, is you're going to get a lot of sexy ladies. Yeah. So again, cool, not too realistic. So this might be a part of the process where we kind of go back to square one in a sense and think like, okay, maybe we need to find reference photos of vintage National Geographic that kind of have this like print aesthetic, but it's of animals to give MidJourney a little bit better information to work with. What I, I want to say is, I love this process of you're just like, finding the right two or three things to mix in MidJourney. And sometimes it's a style reference and a prompt. Sometimes it's a prompt and a mood board. Sometimes it's a mood board and image reference, you know, and so you can just combine all these things and ultimately iterate to a package that you want to, want to send to clients. Yeah. And then I'll just deliver this usually in Figma. The only thing with MidJourney, they don't have any sort of like sharing. I think I've literally seen designers like, if they're on a big enough project, they will like literally create a MidJourney account just for that. And then what I usually do, I'll literally just paste in, in Figma, what that like final prompt is. So the most important stuff is like the profiles that you're using for this one. I happened to go like very crazy with the profiles, stylization if necessary. And then I just say like, these are the reference photos. So for like all of these images that you see here, I think like a hundred percent of them, if not like 90% of them were generated with like this exact setup right here. So I just give the clients this. I'll give them a set of images. Also, these images are going to be like in context. But yeah, it's still kind of a wild west. What I think is really cool about this, and the reason why you're leaning into it, I really appreciate, is, you know, in the past, and we've talked about this a little bit in our episode with Zach at Gamba, who got a very similar package like this from their, their brand team, is in the past, brand and creative directors or agencies would like give you these photos and be like, cool, call us and re-up when you want more photos. And what I love is that you're like, look, I put in all this, you're going to, you're going to value me for all this upfront work that I'm going to do to define the space, give you these codes, like really give you reference images. And then now you can go do this for yourself. And if you want to evolve the brand or you're not quite getting what you want, great. Come back to me and I can give you another package that we can go forward. But it's just like a very different model of providing service. And I think it creates a really positive collaboration between the client and the creative director. I know, sometimes I'm kicking myself. I'm like, gosh, I probably should just not get it to them and just consider charging them. But I'm kind of allergic to retainers. I like, I like doing the beginning process so much that I don't want my old clients to bother me. I just want to keep making new stuff. Well, I, I love it. And I'm sure people are going to see this and you might have a few more for new, new clients. Okay, so you showed us kind of end-to-end how we get through these packages. What are just a couple other workflows that you find yourself using? I guess to continue on with this one, I can show some that I already have done. So for some of the images in MidJourney, I'm sure you have all seen the terrible hands. Oftentimes too, when you're doing kind of like more vintage aesthetics, if it ever can give you an Apple logo, it might give you like some weird old computer. So something I do a lot of the time, even just for my personal stuff I post on X. I'll take my MidJourney images and I'll take them into Flora or Hatesfield sometimes. And I'll just use Nanobanana as Photoshop. Nanobanana literally is just Photoshop. That's exactly how you should think of it. You're just able to speak to Photoshop, essentially for most people. So what I have here is this image that I really like that we generated it, but I want to upscale it to like get more texture in her shirt and stuff. And then I want this to be like a real computer. Nanobanana does require a little bit more prompting than MidJourney, but it's much more, I don't want to say forgiving. It's much less complicated. Like if you're a beginner in some ways, depending on what side of the line you stand on, whether you're more like technical or more artistic. But anyway, I just said, replace the computer she is typing on on a 2026 midnight black MacBook Pro. So Nanobanana is like the reasoning model, so it like actually knows what things are. So you don't have to give it a reference photo all the time, especially for stuff that's in like the public mind sphere. I also usually mentioned like, don't change anything else. I say keep the position and the size of the computer exactly the same. And then just because I've done this so many times, I know sometimes Nanobanana might, you know, change the angle slightly. So I just say exactly, I say exactly what it's seeing. So like only the left side and the keyboard is visible. And yeah, and then if I were to download this photo, it's going to be like 4,000 by 4,000 versus like 800 by 800, kept the style pretty much exactly the same. We just slotted in like a real computer. So it would be relevant to like use now on social media. I'll just know, behind the scenes at How AI, we use a very similar process to upscale screen caps from the podcast for our YouTube thumbnails. And so, you know, I like make, make these faces and then we want to clip them for the podcast. But just screen capping from the video is really low resolution. And so we use a very similar prompt to upscale and improve the lighting on, on our photos. And then we drop those into our thumbnails. How I got to show you, you don't actually have to take thumbnails anymore. Oh, show me live, live demo. Have you been seeing my like X articles? No, show me. Detour. So this is a really like kind of for us, all these like notes and stuff. So my profile photo is AI. I think I got quite a few attention for doing that. I think I ended up making a Hakesfield at some point, but basically what I do is I'll take a bunch of selfies of me that are like more realistic looking. Because sometimes like, you know, you'll take like the best selfie and like, it doesn't look like you at all. Also like show my teeth. I don't have like perfectly straight teeth. I'll take a reference photos and then I can just make me do whatever I want. I love it. This was another one. I had a mid-journey photo that was this, and I wanted her to be annoyed. I kind of already have this process in my head. I knew I wanted a photo of me in this vibe, but angry. So I took this photo, made her angry, and then I replaced my face, which like me on my best day, this a hundred percent looks like me. I have like multiple reference photos to make sure it's like actually getting my vibe. And then from there, I'm able to, you know, I was thinking about doing like an anti-agency post. I think I just ended up using this photo. This was one, took me a couple tries. This is where again, like reference photos kind of really come in. And sometimes, especially with Nanobanana, sometimes you just gotta like write that prompt, you know, like with the actual camera angles with the background. But I ended up just finding these on Pinterest, I believe, of like the aesthetic reference and the actual pose reference. And then this was from my article on AI legal stuff. What else have I done? This is one recently. Funny enough, I could not get it to give me extra fingers, which is hilarious because there were years of my life where I was trying to do the opposite. So for 15 minutes, I was able to get it to give me like AI fingers. That was a recent article I did too. Again, we're like using SREFs. This one didn't actually kind of like carry through, but an earlier references it did. So I just want to call out for people who are listening, not watching, because I'm making a face, which is, Jamie basically dragged in a bunch of like realistic, but still my face, which is really hard as somebody who's trying to generate images of themselves, realistic, but still my face selfies. And then has used Flora to generate a bunch of mix and match, remixed versions for articles and thumbnails. Y'all, my YouTube thumbnails are about to get real, real good. I'm so excited. This is very, very helpful. And if I could throw back to probably three, two or three years ago, so it had to be three years ago. One of the first things that I did was back in the day when you actually didn't have these beautiful UIs, is I fine tuned a model directly on my face that I could model. Like it was like a light, I forget what it was called. Yeah, it's a Laura, I think. Yeah, yeah, a Laura model. So I fine tuned It's meant to be able to make you create consistent client-level work, like I showed today. So join the course. Awesome. We'll link to that in the show notes. Well, Jamie, thank you so much for sharing this. And I'm going to go dive into MidJourney. Awesome. Cool. Thanks, Claire. Thanks so much for watching. If you enjoyed the show, please like and subscribe here on YouTube or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaiPod.com. See you next time.