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 name is a great way to tell the model a ton of stuff without actually having to tell a ton of stuff.

In the past, brand and creative directors or agencies will 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 Claire 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, Nana 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, Ryder, 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/slash. 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 she 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, is the most important part, 'cause I can get a one-off image that's this great, but I can't get this brand portfolio. So you gotta tell me your sorcery.

Yeah, I mean, it all comes down to like, unfortunately having a very tight and manicured process, um, which thankfully I have spent my 10 gajillion hours in Midjourney and Nanobana and everything to figure out exactly what that is. So you're not pulling your hair out. prompting all day. And yeah, I would, 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 1?

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 I thought this was like a really cool aesthetic. to start with. And I typically start— there's 2 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, in the create part of my process, I'm just trying to get information. I'm trying to figure out like what, what are the images telling AI? What is My prompting telling AI what 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 standalone 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 gonna be a better approach to get things going.

One thing I want to call out, if you could go, 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, style references in, in terms of Midjourney, we've done one with Zach, um, 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 lang— 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. And so I love this. A picture is worth a thousand words. Like literally a picture to an LLM is worth a thousand words. Uh, the other thing that I wanna do, if you go to your mood board versus what was generated, one of the things I wanna 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. Um, 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, uh, 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 4 images aren't in the same style as the bottom. And that can sort of give you a seed to start.

I try and avoid prompting at all costs in my process, but, um, like for this example, we're gonna go like super, super simple with like using assets and mood boards. But yeah, if I'm in one of these, like I'm doing some insane editorial work for clients, it needs to be consistent across like 100 images and like foolproof. to like a consumer eye, that's when like getting really into the details, especially with models like Nano 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 5 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 can 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 ended up doing is I just removed that green eye SREF. This is something that comes with intuition. You know, I've been using Midjourney for like 3 years, so I kind of can like understand like where— I don't know, it feels like sorcery to me, but I can understand like where the LLM is pulling certain things. And I knew that if I removed this green, it would solve a lot of my problems. Um, so I did. And as you can see, we're starting to get a little bit more neutral in tone. We're also like a little bit more zoomed out too, especially with like the people photos. So now I know I am On the right direction.

Can I ask a question real quick? When you say I'm using the SREFs, where are you getting those from your mood board, or how are you actually deciding what the SREFs are? Just for people who are less familiar with Midjourney.

I'm just using literally the ones that were on that original Pinterest mood board. So you can just copy and paste that image and put it in there. And then as you copy and paste things in, it'll like save in a library for you. Like forever. So if I wanted to bring that green one back in to keep trying stuff, it's like already there for me.

Got it. So instead of the SREF codes that a lot of people, and we've talked about in the past, you're using literally the UI to just drag in images as style references. And that for you sometimes gets you better results than just using the general mood board process.

Yeah.

Yeah.

Cool.

And then I have one more question, which is, do you have some go-to like test prompts? Like, I love the astronaut as a prompt, um, because there's a lot of ways you could generate an astronaut. Do you have some like go-tos that you run through when you're doing a mood board, or does it really depend on your client and what you're working on?

Yeah, I would say I love doing like ethereal female model for some reason. I do like— I do that a lot. Um, yeah, this is like a crazy silver one that I was doing. It tends to just like give sort of like a more elevated vibe than just regular model. I think I do cats a lot because there's just a lot of like texture to work with.

The other thing I would say is that there is probably a lot of training data of cat pictures on the internet. So Midjourney can probably do a pretty good job with a cat.

Yeah. Running too. I feel like I do that a lot. Oh God, this is like my first generation. Isn't that crazy? Um, I don't know. I scrolled all the way down there, but, uh, yeah, I'll do running a lot or like runner. Yeah. Anything with like, astronaut's fun too. Anything that's like specific enough to kind of like give you the vibe. I hate saying like vibe. It's like kind of nonsense.

But broad enough that those different styles kind of apply. Okay. So, so you've used these style references, you're getting a little closer. What's the next step?

Yeah, so I consider it as part of my like create step. And again, the goal is just to like get some like information. Take like, this should take like 10 or 15 minutes to start the create process. And then what I'm going to do next is I'm going to move on to iterating. And that's when we're going to start like getting a little bit more specific with our process, maybe slightly more technical, and then starting to combine more styles to get like your own unique style and get better consistency. Um, so what I'm going to do particularly in this process is I'm going to start to bring in some personalization codes and other mood boards. So what ended up helping me get these results is my personalization code that I call Late 2025 Aesthetic. Um, now personalization codes, again, it's kind of like a little bit of a mystery how it works exactly under the hood, but when you're creating a personalization code, Midjourney is going to put you through this like endless flop matrix of images that you're either going to vote 1 or 2 or skip. Um, and basically it— you're just telling it what you like, and you can have like as many profiles as you want. So for like this profile in particular, I was trying to think— I was trying to rate images that were of a 2025 aesthetic, so like more iPhone style. And we can see here, like, you'd pick that one. Yeah, or I, I skip a lot. Um, that's another thing when I'm giving like advice to people on personalization codes is like there's a lot out there on You know, like Midjourney says, like, don't skip that much, but some people are like, skip as much as possible. Some people like skip a medium amount. I tend to skip like a medium amount and only pick things that I would like if I generated it. But I do find there's kind of like style bleeding. So for example, let's say I like the quality and the colors of this image. If I like a bunch of images that look like this, I might get a heavily influenced style that like wants to be painted.

Yeah.

Or wants to be like this like vintage or Renaissance, whatever aesthetic. So that's something to consider as you're going through this.

One thing I want to call out for folks that are— it's a little bit of a side of this, this particular flow, which is when you're building an AI tool yourself, I've— I have not seen this like this or that personalization flow in a lot of AI tools. And I think it's such a good way to like fine-tune whatever you're going to provide to your end user. And so a lot of times we get these like end A/B tests and like a ChatGPT prompt. But this is so interesting to kind of put this up front and say, okay, let's spend 5 minutes telling me what you like. Then we can be more confident that when you have a downstream experience in my, in my AI tool, it's gonna look great. Okay. And you can create as many. I've never, again, I have not used mood boards or personalized. I'm just like YOLO up in the, in the main chat. So, and you can create as many of these as you want.

Yeah. Yeah.

Um, to, to use. Cool.

Yeah. Unfortunately, you can't like go in and like edit them the same way that you can with mood boards, which is why you gotta be good at like naming and trying to understand like what you were doing 6 months ago when you spent 2 hours ranking images. Um, but I have my own crazy system. That's just a quick brief on like what personation codes are. So what I started to do, let me go back to these like other images. I, I just felt like these could use more of my own style, and I felt that they were a little bit over-stylized. Like, I do enjoy the pink, but I was curious to see if like maybe we can get some like other influences in there. So by adding my personalization code, I was able to just get like more of a depth, I would say. Also because I wanted it to be like very crisp and like modern, you know, um, you start to see like Better skin.

Yeah, this one to me looks like very, uh, a good match for some of your earlier ones. That one's the one that stands out to me.

I mean, it's kind of, it's, it's nuanced, but 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.

Yeah. 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. Yeah. So how do we take this to the next step?

Yeah, so I want to show you guys some like of the other Midjourney techniques and ways, um, 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, um, but it's Dazed editorial photo shoot. So Dazed is 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, it's like a really famous like ASAP Rocky cover. Um, it's like super gritty, high contrast, all these words that like are really hard to find. even if you are like a professional photographer. But when you just say Dazed editorial, depending on how like famous the publication is, Midjourney's gonna know what you're talking about. Or, you know, Vogue especially, it's gonna know kind of like the level of the highlights and okay, we're gonna be doing fashion stuff. So I love doing that. Same thing with the word editorial. In this case, I say a woman in her mid-20s, deep in thought, up close macro photo. So not the most crazy prompt. I think everybody knows what a macro photo is. Up close, it's not like, you know, center frame, zoomed in. It's very human language that I'm using. So what I did for this prompt is I wanted it to really have the vibe of this original S-ref we have here. I wanted to make sure it was like the same composition. So what I did is I used an image reference. Image references are kind of tricky because definitionally they just structure your composition, but a lot of the times like the structure of an image and the composition kind of is the style. So I kind of don't like to separate style references and image references. If you have a woman posing like really sexily and you use that as an image prompt, that stylistically is going to influence a lot of what's happening in your image. It's going to immediately be more editorial or sensual, etc. But anyway, I use this as an image reference because I want that profile. But obviously, as you can tell, We're getting bubblegum immediately, which I don't hate some of these, but my intention was more of like a thinking photo. So what I did is I literally just zoomed in, like in the Midjourney UI, I cropped it, and then I just like dragged it back in. And then I ran the prompt again, and then we got something very similar to what we were working with here, especially in terms of like image quality, the colors. For better or worse, it's also kind of like mimicking her like bone structure, um, especially this one. You can see like we have like these contours. So image references will give you a lot in terms of style.

One thing I want to call out that I've noticed in 2 of your flows is on the we're getting too much green, there was just such a compelling element in that green eyeshadow photo, which was like half of the eye was just very green. It was clearly the most obvious thing about the photo. And with this image, the most obvious thing about the photo is she's blowing a bubble of gum. And so what I like is in both of those, you're like, just boot the thing that is so obvious and so overwhelming. And you can either do that by kicking out the image, or you could do that by cropping out the part of the image that is pulling the rest of the generations down. So that's a really clever technique. And I love that you're just like, You're like, I'm just gonna screenshot and drag it over. I'm not gonna like try to prompt it to say like remove the bubble gum or whatever.

Yeah. Oh my God, you will spend all day in Midjourney, you know. Even if you're using like— something you can do is like the no feature. You can say no bubble gum, um, you know. You can try and like say like lips are showing. Sometimes it'll work, most the time it won't. If you really want to get specific, you could take this into Nano Banana and say like remove her bubble gum. But I don't think it's gonna get us—

well, that's what I was gonna say. And I know we're gonna get to Nano Banana in a minute, but then also if you move into Nano Banana, you are gonna wait like 45 minutes for that photo to come back. Yeah. And like waste a bunch of credits and stuff that you just don't need, right?

Yeah.

To do this.

Yeah. That's another thing too. Like avoid like doing all this SREF and this, like this, um, the profile codes. I, I don't have to prompt like 2025 aesthetic. Uh, you know, like, model-esque. Like, it already knows that I'm gonna have like a very conventionally attractive blonde woman because I gave it a conventionally attractive blonde woman, um, or vice versa. I don't have to type in the colors and the grading every time. I don't have to do those horrible JSON prompts that no one can use. Um, so this is like the laziest way ever.

I love it.

To do prompting. Okay, let's go into some like more specifically prompting, my arch nemesis. So one thing I wanted to do, again, I like the juxtaposition of like the images on the original mood board. I thought it might be fun to have like a deer in a New York City apartment, kind of like bringing like the forest into the city. That could be cute. I got these images, like they're definitely on par in terms of like general style, but I was kind of missing like the New York aspect of it. So instead of like going and maybe finding like a New York City SRF or like finding an image of an apartment putting it as an image reference, which potentially could have worked. This is one time that I do actually do some prompting. So all I said was like, New York skyline can be seen in the window behind it. So again, not doing anything super crazy with the language. I'm literally just saying what I want to see. Another example, I thought the pink couch was a little bit too fantastical. So instead I just said on a matte black Leather couch. And now we're starting to get— we know it's in New York. It's a little bit more kind of like realistic, I guess. A couple other things I tried was, uh, I said at night. I think it's kind of giving me at night, maybe sunset vibes. And then one, the craziest I'll ever get is I have a list, uh, this is included in my course, of like all these different cameras. Um, so I have like DSLR, I have mirrorless, I have digital, I have film. As kind of like quick shortcuts, because that's really hard to remember all 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 legs 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 to know with the prompt is like saying luxury New York City apartment instead of saying on the 4th 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 not LLM, the AI knows luxury is gonna 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 Midjourney 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 Midjourney and you look through people's prompts, like, people need a job, man. Like, this is too long. And so I love the idea of 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, I nail things really fast, like Like this aesthetic I happened to get like very quickly. Um, 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 prompt do I have here? 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 SRF, which actually came from a previous Midjourney generation I did that day. And then we get all this cool stuff. I think, yeah, the prompt's experiencing music. So we're starting to see ears and maybe body sensations. That's another tip too, is 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 to get the model thinking. Same thing, like I'll just do like financial markets, really vague stuff. Midjourney is like very poetic, I would say. You can like write little poems to it and it'll actually give you some, it'll do what you want it to do. Like, this is experiencing music to me, you know what I mean? You feel like the vibrations through your body versus, you know, if you were to describe this image, it's gonna say like a whole bunch of nonsense.

Even worse.

Even worse if you do it in like Gemini or something and ask it to give you, you're just never gonna be able to get this, which is why I love Midjourney so much, 'cause it just feels like an extension 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. Um, And so I love the idea of just getting out of like the tactical and practical prompting of like, I want this thing in this camera, 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. That's lovable.dev. Okay, so just to recap really quickly on our Midjourney journey, we have done mood boards. We have used those mood boards to kind of, um, 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, um, 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, dear 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 SREF 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 into the dreaded like 5-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 hugely 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 in this 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. Um, 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. Um, 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. Um, so what should we prompt? You're like turtle in the sea.

I was thinking turtle. This is weird.

You're thinking turtle?

We're floating through Midjourney together. I was like, maybe she'll do like a turtle.

Yeah, very funny. So this is like one again, kind of back to the beginning. We're gonna try the mood board, see if it works. If it doesn't, um, the good thing is we already know this SREF stuff works. Something else that I was trying, um, I'll take that new mood board with the images And then I'll try with a different mood board that I have. So this one's called Real Skin. This is just when I'm trying to make like really realistic images of skin or of models. Obviously it influences a little bit of like what they look like too. So I will use both mood boards at the top at the same time. I might just say model, and then we're able to kind of like play with using Rose Aesthetics. Um, now I will say, so these are generated with the mood board, definitely like closer than like if we hadn't used it, but I still think the SREF is gonna like be superior here. Yeah, so what I'll do is I'll just take that, um, this mood board here, and I'll click that, run the same prompt again, but using the SREF and using the mood board. And we'll see what it gives us. Yeah, so same for this, like closer to what we want, but like not as like stylistically consistent as these, which sometimes depending on what you're making, like in my opinion, like all these images are great, but like over time it might start to look a little bit flat or like too much pink. So like weaving in a couple images that are like not the exact same style can also be good too, as long as you have this kind of like through line. So like a good example here is like This turtle, our mood board and our SREFs is like not super optimized for animals.

Yeah, that's what I was gonna say.

Yeah, it's way more optimized for like editorial stuff. So what I might do here, I do have like a nature mood board that I have. Adventurecore. Um, and I never delete my mood boards. So I'll just like keep them around and we'll see. Yeah, we'll see what this gives us, um, using the SREFs. I already see we're getting a sexy lady. Um, naturally.

That is one of the downsides of Midjourney is you're gonna 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 2 or 3 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, um, 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 happen 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 100% 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 the 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 Gamma, 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 gonna, you're gonna value me for all this upfront work that I'm gonna 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 wanna involve the brand or you're not quite getting what you want, great, come back to me and I can give you another package and we can go forward.

Yeah.

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, and the creative director.

I know sometimes I'm kicking myself. I'm like, gosh, I probably should just not get into some and just continue charging them. But I'm kind of allergic to retainers. I like, I like doing this 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 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 Higgsfield sometimes, and I'll just use Nano Banana as Photoshop. Nano Banana 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 wanna upscale it so I get more texture in her shirt and stuff. And then I want this to be like a real computer. Nano Banana does require a little bit more prompting. than Midjourney, but it's much more, I don't wanna 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's typing on, on a 2026 midnight black MacBook Pro. So Nano Banana is like a 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 mention 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 Sandov and Anna might, you know, change the angle slightly.

Yeah, zoom in, zoom out.

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 gonna be like 4,000 by 4,000 versus like 800 by 800. Kept the style pretty much exactly the same. We just saw it in like a real computer, so it would be relevant to like use now on social media.

I, I'm just gonna— behind the scenes at How I AI, we use a very similar process to upscale screencaps 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.

Oh, I got to show you. Um, you don't actually have to take thumbnails anymore. Oh, oh, okay.

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, uh, my profile photo is AI. I think I got quite a bit of attention for doing that. I think I ended up making it Higgsfield at some point. But basically what I do is I'll take a bunch of selfies of me that are like more realistic looking, um, because sometimes like, you know, you'll take like the best selfie and like it doesn't look like you at all. I'll also like show my teeth. I don't have like perfectly straight teeth. I'll take a couple reference photos and then I can just make me do whatever I want.

I love it.

This was another one. I had a Midjourney photo that was this, and I wanted her to be annoyed. Uh, I, I kind of already had 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 100% 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 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 Nana Banana, sometimes you just gotta like write that prompt, uh, 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 for 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 'cause there were years of my life where I was trying to do the opposite. So I was like, 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 in earlier references it did.

So I just wanna call it for people who are listening, not watching, 'cause I'm making a face, which is Phoebe 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.

Yeah, real good.

I'm so excited. This is very, very helpful. And if I could throw back to probably 3, 2 or 3 years ago, so it had to be 3 years ago. One of the first things that I did, um, was back in the day when you actually didn't have these beautiful UIs, is I fine-tuned a model, uh, directly on my face that I could—

In the LoRA?

Like a, it was like a light, I forget what it was called.

Yeah, it's a LoRA, I think.

Yeah, yeah, the LoRA model. So I, I fine-tuned it on my face and then, um, I generated a bunch, a bunch of images and it was so useful to have this specific one. And now we're all spoiled. We can just do this in this UI. And I never wanna see this, um, Chad Steve Jobs ever again, but I love the idea of dragging these 2 nodes together, um, into, into an image.

Yeah, people, people struggle a lot with it. It's hard to. To get to nail it. I think it's— I have like a really kind of like strange face. I have like big eyes and stuff that like just latches onto me really well. Like, I have an Asian boyfriend and it just like refuses. Like, every generation I make of him is just like generic Asian guy.

We could— we can do the aside of like the— like the bias of what goes into these training, training sets. I mean, we can get a lot of conventional, as you said, conventionally attractive blonde women.

Yeah.

We can generate all day, all night, 'cause they're all over the internet. Um, fingers crossed.

Every, every athlete, if you, if you want an athlete in Midjourney, it's gonna be a very large Black man. It's going to be a very strong, beautiful Black man.

Oh my God.

You cannot get a white, you can't, you cannot get a white athlete in Midjourney.

I mean, this is what's really interesting is, is you start to play with these models, you really start to understand they are trained on the internet.

Yeah.

And I was mentioning recently in some like chitchat about, um, Moltbook, which is where all the open claw lobsters are together. And people are like, why are they weird? And I was like, they're weird because it's a fake Reddit trained on Reddit data.

Yeah.

And people on Reddit are weird. And so I think one of the things you have to know about using these models is like, where does the training data come from? And therefore what's it gonna generate for you and what isn't? And I, and I, the more you explore it, the more you, more you hit those edges. Jamie, I love all this. This is all super useful. I'm gonna be able to go put this into practice. literally today because we're working on a thumbnail today.

Awesome, yeah, today prompts.

So let's, let's skip over to lightning round and then I will get you out to generating another 1,000 images. So my first question is, where do you go for inspiration? I think one of the things that you sort of like pass through is you're like, oh, if you know this magazine or that editorial style, that camera. And I feel like people aren't cultivating their visual taste and language enough. So what are some of your, you know, other than Pinterest, what are some of your sources where you're Um, continually improving your language and your aesthetic taste through exposure to other things.

Yeah. Um, one thing that I recommend to everyone is to start a list on Twitter. There's tons of like basically Tumblr accounts still to this day that post like really aesthetic images. So like some of these are like, uh, this account's literally called Pink Glitter. So she posts a lot of like 2000s stuff. Um, so that's really great to use for Midjourney. Fashion accounts, a lot of fashion accounts who I follow. So I have like a Twitter list of like really aesthetic photos. Like this is obviously great if you again want more sexy ladies. So that's what I do. Also Cosmos, of course, is superb. So let's do like the real model here. One of your favorite images. Yeah. Um, so obviously this is just like a treasure trove of ass drafts, especially for like specific moods and vibes. I feel like as a designer, I still, as much as I love Cosmos for like branding stuff, like logos and like website inspiration, I still use Pinterest a ton as like my home base. But I will still use Cosmos for like art direction things. The great thing about both of these is like Pinterest has a plugin, um, so you can literally save from like any page, especially like again, if you're on like X, you can go save to Pinterest. Cosmos, I believe, has the same thing, but I don't have it turned on right now. So that's what I do. And then anytime I'm like shopping online, again, X, Instagram, um, something I teach in my course as well is a daily taste practice. So just like get really on top of saving and archiving stuff. My mood boards on Pinterest are actually shockingly simple because I realized if I had too many, I was like uh, getting kind of like decision fatigue, which is why my design inspiration board is 7,000 pins. It's probably time to like switch that up. But I have like, I don't know, what is this, like 20 mood boards? So like instead of having like ethereal model mood board with like 10 images, I just have like one that's like Sref. And anytime I see anything on Pinterest that I would like love to bring in to Midjourney one day, I just save it. And then when I need inspiration, For like any project, even like client projects. Like I'm doing a sock brand soon, so I have a lot of running stuff. I just save it all here and then it's really easy for me to go through and just like copy and paste into mood boards. As you can see here, like these are images that I've saved. I don't know what the hell I was looking at.

You were looking at that, like it's like shells and it seems like very similar.

It was like a website for like technical 3D scans. But this is like a grain of rice or something, but it was sick. So I like saved this and then I really need someone to do this for me, but I get like millions of views on Pinterest a month now because every time I see something and I save it, like images I was already gonna save anyway, like this is just from like high fashion Twitter. I just saved it and it has like 200,000 impressions. So.

Not only this, this is a kind of like source of inspiration for you, it could actually be a source of business for you cuz you're sharing out and making these resources useful to people. Jamie, uh, my last question I have to ask you, ask everybody, which is, you know, you, you have seemed to— your prompting technique is keep it dead simple. We also talked a little bit before the show about how you don't love to prompt a lot and you're not like, you know, you don't wanna spend a bunch of time in Claude. But yeah, if AI is not giving what we want, if we're getting the ugly stuff here, what's your prompt? What's your prompting or personal technique to just get it going in the right direction, whether it's images or text? Are you mean?

Firstly, take a break. Always. Like, you're never going to be able to, like, see properly if you're kind of, like, in it. Um, so, like, sleeping on it or just, like, walking away. And then sometimes I come back and I just, like, new SREF, different prompt, like, grab this reference, use this camera, and then it immediately works. Um, so I would say walking away. And then part of my process too is just like during that walk away, when you come back, be like, okay, what is actually the problem? Like for some of these generations, I'm like, it's just too busy. It's too Midjourney. I'm using like 18 SREFs. Like this is just not gonna work.

Yeah.

So I'm gonna like take a step back. I'm gonna redo the mood board and I'm gonna figure out what exactly from these images is important. And it's like, all right, I keep getting this like stupid red color. I'm gonna take out this red color. Like even though I love this image, And I would love for, I would love for it to like come through. It's not giving me what I want. Midjourney's not listening to me. So again, like, I think I talked about this in the beginning, like brutal honesty on like what's going on and trying to like not, I don't know, make AI work the way you want it to work and actually understanding how it works in general. But I think time away is probably—

I love it. Time away and seeing things from the AI's point of view.

Yeah.

Well, Jamie, this was great. How can we find you and where can we be helpful?

Yeah, um, so I spend all my time on X. So, uh, X @JamieGannon or TechBimbo is my name. And yeah, I have an AI course coming out. It's called, um, The AI Creative Director. It's on Maven. Um, you can find that on my X or my website as well. So if you are interested in like really deep diving into this and getting live coaching from me and hearing more of my by yapping on AI, I would highly recommend you do it. Um, 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 gonna go dive into Midjourney.

Awesome. Cool. Thanks, Claire.

Thanks so much for watching. If you enjoyed this 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.