Leveraging AI

96 | Create stunning and engaging graphics aligned with company branding and campaign style guides - Master AI visual assets with Drew Brucker

• Isar Meitis, Drew Brucker • Season 1 • Episode 96

Share episode



How can AI transform your business imagery effortlessly?

The idea of using AI to generate consistent, high-quality images might seem daunting. After all, aligning AI-generated visuals with your brand’s unique style and requirements has been a challenge.

But what if you could harness AI to create a cohesive visual identity that enhances your brand, all while saving time and resources?

In this episode of Leveraging AI, Drew Brucker, a digital marketing ninja and AI imagery expert, shares practical techniques to master MidJourney, the leading AI tool for generating consistent, brand-aligned images.

-----------------------------------------------------------------------------------------------
Don't Miss The Biggest AI Event of The Year - LEVERAGING AI Ep. 100 LIVE!
20 top AI experts, each sharing their no. 1 tip for AI business efficiency.
Join us on Monday, June 17th, 12-2PM EST
Only 100 spots are available - sign up for the event now!
------------------------------------------------------------------------------------------------

In this session, you'll discover:
- Why consistency in AI-generated imagery is crucial for brand identity.
- How to create and use style references, image weights, and character references in MidJourney.
- Practical steps to build a library of custom, on-brand images effortlessly.
- Tips for troubleshooting and refining AI-generated visuals to perfection.

Drew Brucker is a digital marketing expert and founder of Brainchild, a consultancy specializing in AI-driven branding and marketing strategies. Known for his advanced use of AI in creating branding and visual assets, Drew brings years of experience and innovative techniques to the table.

Here is Drew's  Style code doc: https://dbruck.notion.site/Midjourney-Style-Codes-80380b08ea0944a3929df260a14d8f50?pvs=4

Example prompt template for brand: 
[photography type], [camera angle] of [ICP/Job Title], dressed in [attire] [action], [background] —ar [x:x] —s [x]

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Isar:

Hello and welcome to Leveraging AI, the podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business and advance your career. This is Issar Matis, your host. And a thing that I get talk to a topic that a lot of people approach me about is, yes, these AI tools are really cool and they're amazing and they're really fun toys, but they're just not good enough for business yet. And that is true, probably All across the board, as far as these people approaching me with that, whether it's not accurate enough to do data analysis, it's not consistent enough when writing in our style, it's not and so on and so forth. And a big one is image generation. A lot of people saying yes, even the people who know how to use Midjourney or Dall-e or any other tools feel that they can create cool images. But they're not aligned with their style guide. They're not exactly what they need for the product. They cannot create consistent images to create a campaign out of it, which hence they feel that while this is being a really cool technology, it's just not there yet from a business perspective. And I'm happy to tell you that's not true. Maybe somewhat true, but not totally true. And there's still a lot you can do for your business within these tools. And today's focus is going to be how to create your own style within Midjourney. So you can have consistent style guide aligned with the needs of your business, aligned with your business style, aligned with a specific campaign you want to create. So we're going to dive deep into how to use Midjourney in order to create consistent, Images. Now we're going to be doing this over a podcast. So we're going to explain exactly everything that we're doing. We're also going to share this on our YouTube channel. So there's going to be a link in the show notes. If you want to see exactly what we're sharing on the screen, you can do both. So if you're driving the car, keep on listening. If you're walking the dog, keep on walking your pooch. But if you are. in front of your computer or something like that, then switch to, switch to YouTube. You will learn a lot more that way. Our guest today is Drew Brucker. It's his second time on this podcast. He's only the second person that comes on twice. And it is because Drew is a digital marketing Ninja. He's been doing digital marketing for many years. And in the past year or so under his own consultancy called brainchild. And in that consultancy, he helps Brands. Develop a better branding and marketing strategies, and he's probably from an AI usage, one of the most advanced people to use AI in the creation of branding and ads and visual assets for companies. And hence, he's the perfect person to lead this discussion for us. Drew, I'm really excited to have you back. Welcome to Leveraging AI.

Drew:

Yeah, man. Great to be back. So much has changed since the last time we talked.

Isar:

Oh my God. it's like a lifetime of AI stuff. Like we talked about probably about a year ago. I was going to

Drew:

say, yeah, I was thinking it's probably been, yeah, about a year. And you just normally that's not that long a time. But in, in AI terms, the compounding of the changes is just insane. So

Isar:

yeah. it's, every. Every day, if you look overall, but if you look through a very narrow lens of, okay, let's focus on Midjourney, every quarter or so something big changes that is dramatic. And I think these two aspects of being able to create style guides and be able to create consistent characters are maybe not the entire Holy Grail, but it's a very big step in the direction of the Holy Grail, actually using this for business. Yeah, big

Drew:

step, something that I think if you've either played with image gen in AI or you've thought about it, it was one of the things that people have just been waiting for. And so that's happened. It's, I would just say maybe it's a milestone moment in terms of some of the things that have launched and deployed, with that technology, but yeah, it's absolutely essential. And I think, One thing that I do want to preface is if you've got any sort of role in marketing or branding, there's a use case for this for you. so just maybe keep an open mind and think about things a little bit differently. And hopefully if you're listening to a podcast in the first place, you're looking for maybe some validation, but also some things that you didn't know or some new ideas that you want to try out. And so hopefully we can uncover and unpack some of those things today.

Isar:

Yeah. Awesome. I think the best thing would be to, first of all, talk about what it is, what is like style reference, what does it do, and then what is consistent character and what does it do? And then we can dive into the actual how to do it within Midjourney.

Drew:

Yeah. style reference I think, is a, is one of those interesting features that. Maybe isn't as critical as something like consistent character, but something that, was probably pretty hard to achieve before in terms of consistency with the style, right? So what I mean by that is, if I'm maybe backing up and thinking about a brand in general and the color palette that they have in their brand, certainly some of those things can be, identified with words, right? you could say black, but there are different shades of black, right? Yellow, different shades of yellow. and some of those things, you can accomplish just through the text, right? So maybe it's a canary yellow or it's a desaturated yellow, but it's still a little bit difficult because some of those things aren't consistent, right? And so a canary yellow on one image may look a little bit different than a canary yellow on another, what style reference allows you to do. and this is just as a whole, rather than specifically Midjourney is just To take an existing aesthetic style and color palette from an existing image and transfer that over. To a new image that you create. And so obviously that's pretty powerful, pretty self explanatory. If you're trying to dial in some customization and specification for your brand. And I think, that's really one of the major unlocks with AI visuals is the ability to promote resonance with your audience. and so how do you do that? it's hard to do that with stock photography. And if you've got a campaign or something and you don't have big budgets, right? Like maybe you don't have the spend or the time to go out and hire photographers for it. And so things like this allow you to get very specific to your brand. If you've got a very unique audience, right? So maybe it's, a very specific niche or demographic. So it could be, maybe it's older males that are 65 and have a tattoo, sleeve down their arm, or it could be, young females, maybe they're teenagers, And they're into this or that. so how do you get those images without being able to custom, to customly create those things and before just so many obstacles there, this unlocks that. And then obviously you bring in things like, The colors, the consistency that makes your brand you, and all of a sudden, you're not only creating unique images, but you're creating images that resonate with your particular audience. They can see themselves in your ads, in your campaigns. And I think at the end of the day, that's super important.

Isar:

I love that. I want to summarize quickly what you said. The first thing is obviously when you're doing any kind of visual campaign, the first goal is to connect with your target audience and to understand that you've got to understand what's going to appeal to them. And this, it, at the end of the day, this is what defines the style. This could be cartoonish. This could be lively. This could be dark. This could be like, whatever it is that you try to create, as far as the feel of the visuals of the campaign, is the hardest thing to create in the real world because Again, you need photographers. You need the environment. Let's say you're trying to do a campaign that talks about shooting stuff in dark alleys. guess what? You have to block an alley. You have to get licenses for that. You got to get a photography through lighting, like all of that, which most companies don't even have the budget to do. But even if you do, then, okay, you're going to shoot 600 photos and you're going to pick from them. And. In Midjourney, if you know how to create styles and use them, you can create 6, 000 of these in a day and finesse them to exactly what you need with less than one 10th of 1 percent of the budget that it costs you to do the other thing that just 100%, 100%. And so it allows you going back to what you said. To achieve what you're trying to do, which is to connect with your audience on an emotional level, because you have a very clear thing in mind of how to connect with them.

Drew:

one of the things that stood out for me is that wasn't just a thought too. I went through that myself, right? I was a marketing leader on a small marketing team, at a startup, you have a budget, right? But it's not. You don't have the budget of a larger company and you have a small team, right? So everything is competing against your time and your priorities. And so while a brand is important, super important in a startup environment, it can get a little bit lost in the grand scheme of things. It can deprioritize. And I'm not saying that's the right thing. In fact, I would say it's not. But that's just the reality. Things are more important in terms of getting revenue off the board, showing wins to your VC or whoever's funding, the startup, things like that. And so this, was a problem for us because we were in a niche space. We didn't have great photography that existed in stock. And if we did, you got to go through pages to try to find something that kind of feels like you. And so that was essentially the same time I was playing around with Midjourney. It was fun. The quality got good enough. And I just thought to myself, I think there are some things that we can use this for in the day to day. And so the idea that I had was really just creating a library of custom. On brand images that my entire marketing team could use. And then eventually, The entire company could use and just grab things without having to go searching for it. And so that is a little bit of upfront work to do the experimentation and R and D to get there. But at the same time, once you're there, you can just run the same prompts in perpetuity. You can run same variations, right? So there's some static things of about a prompt, but there are some things that are dynamic, like the age of. your audience or what they're wearing or what they're doing or where their location is. And so you just think about all those interchangeable things. And now all of a sudden, anybody from your company can grab something for a website, a social post, an ad external presentation, internal presentation, right? The use cases are endless. And so that inflection point for me was critical because I just thought to myself, I know other people are in the same position. I know that for a fact. And if they care anything about brand, this is a pain point, right? And so that's where the technology is at. Like you said, at the beginning of the show, there's still some limitations and what those things can do at the same time. this is something that you could immediately start putting to use right now. Plenty of use cases, in plenty of like non risk associated with some of these things too.

Isar:

Awesome. Great segue. Great setup. Let's dive in. Let's see how actually it's done. Like, how do I create a style in Midjourney and how can I reuse it to create the same vibe, the same thing, the same branding across multiple images.

Drew:

Yeah. So let's, I'm going to share my screen here. I'm going to walk you through one thing really quickly in terms of what some of those identifiers are. And then I'll just dive in and show a real, maybe example here. Awesome. What I, for

Isar:

those of you who are listening, we're going to walk you through what's on the screen. what we're talking about, and we're going to. Describe as much as possible how to do this. And for those of you who are watching, then you're watching. There's no problem.

Drew:

Yes. great call out. So this is just a slide deck I've put together to showcase how I've thought about this in the past. I think some things have changed, but the basics are really the basics they haven't changed. So what I mean by that is. Who, what's your brand style guides as they, they are today. What's that style. And also thinking about that one step further, is that what you want it to be? Because now you have. Some creative compromises that you don't have to make anymore, right? So that's important as well, just because your brand is what it is today. Are there any sort of evolutions to that, that now are unlocked because of something like this? So what's your brand identity? What's the style guide? And then really, thinking about the things that make your brand new. and what I mean by that is going back to the target audience that you have and putting yourself in the shoes of those people. What do they wear? What do they look like? where do they find themselves working? what kind of expressions do they have? and all of those sort of intangibles with your brand, it's important to note those things. And then we can build something that's very stylized and custom to the brand. And so I always back up first and think about what those visual brand descriptors are. We'll try to, what kind of feelings am I trying to drive? what are some adjectives that describe my brand? So some of those components, and this is just an example of that, When I went through this, I did this myself. I thought about all the things that made our brand, so it's a tight knit community. they're undervalued in the community, Overlooked. normally they work in darker environments, like it's moody atmospheres, And so like really identifying those things in advance. Again, going to the target audience. And these are just some examples and this is previous version images too, right? So where we're at today, the images are even more realistic looking than these, but you can just see the theme here, right? these images were all created to have a very specific look and feel, right? They're moody. They incorporate the brand colors, which are brand, black and yellow. you've got actual residents with the people that are in them. This is what they wear on the job. These are the type of environments they find themselves in. This is the type of equipment they use, right? So I took a step back first and identified all of these things, right? So I think that's important to do. A lot of this can probably be fine found in your brand identity or style guidelines. so that's not a big deal there, just really detailing those things and then thinking about what type of Atmosphere you're trying to create, like what type of library of images are you trying to create for your brand? How versatile are those things? So I try to think of, Hey, my personas, I have older males, middle aged males. Middle aged females, young females, right? These are their ethnicities. These are the things that they wear. And so interchanging all of those things. So I've built up this library where I've got images of guys, older guys, younger guys, middle aged women, young women, right? different ethnicities, white, black, Asian, whatever that is for my particular audience. And now I can just grab those whenever I want.

Isar:

And so I will add something and then I'll ask a question. I think that step is actually something a lot of people miss because especially people in a marketing, a marketing part of the company, whether it's a department or that's part of your role or whatever the case may be, the usually have your day to day and it's really stressed and you don't have enough resources or enough time to do the stuff that you need. And the blanket is really short and you have to do all that stuff and you trust the, okay, this is our guidelines. That's what I'm going to use. And I think what you're saying here is very important by the fact that it forces you to go back for a second. Why is this the brand guidelines? Who am I trying to connect with? Is the fact that we defined this five, six, 10 years ago, still accurate. Yeah, in the pandemic sense, there's been a generational change of the workforce. There's been AI show like maybe we're not serving the same exact target audience as we did before, or maybe the same target audience has slightly different needs or it's slightly different people and so on. And so

Drew:

things have changed or evolved or, I think going back is guilty of this, Where they haven't updated these things in a while. And employees may come in and out in the guidelines are exactly the same as they were before.

Isar:

So I think going back and reflecting and doing that exercise is very important. But here's a question for you that I think we will probably answer as we dive deeper into the technical practical side of this. So you said what you did, you said, okay, I know Midjourney really well. I will create this. database of images that anybody in the business can use. And I think that's awesome kind of creating your internal library. But I think there's an option, especially today with styref and with, with consistent characters to maybe even create a prompt library where you can give people that structure of a prompt and they can create their own. Images as needed.

Drew:

That's right. That's right. you're predicting the future here. so so this is all sort of a preface to that point, right? one other thing I would just touch on before getting to that point is references, right? Mood boards, like where's your inspiration for things that you want to create, right? And grab those images. They could be existing or maybe they're just other places across the web, third party sites. That's okay. Or you can even create those things in Midjourney. And perhaps we can even show you that today. But the point is bring in some visual identifiers to help you there. And so now you've got all these things to work with. You've got all the materials, the raw materials to put something together. And then. This is exactly what I was getting into creating these prompt templates, right? I would just call them like prompt templates and then a prompt library. And I've done this for a few brands, which is essentially the goal is to get to some place where you've got static parts of the prompt, right? See things that maybe don't change so much. And then you've got things that are dynamic, right? Things that you can swap in and out. So those would be like ages, ethnicities, attire, locations, actions, facial expressions. But none of that really impacts per se, the colors of the brand palette. so that's maybe where more style reference fits. And then now you're just focused on the dynamic parts. And so what type of prompt or prompt structure is going to work for you, so I think that's something that comes into this experimentation R and D category. There's not necessarily a universal thing that I've necessarily found because, Midjourney's database know some things really well and it doesn't know some things very well. And so some of that, NLP or the language within it. needs to be tweaked and I'll give you a perfect example, right? Like I've got a client that is in the metal recycling space and some of the equipment, for example, like a, like one of those metal claws that picks up all the scrap, it doesn't know what to call that. So there are like five different names for it. I go through all the synonyms trying to describe this to Midjourney. It still doesn't know what it is because it hasn't trained on that much regarding that. so then how do you troubleshoot it? And so those are some of the things that you have to think about, but this is the goal, this is the goal, because once you have something like this, you can just run these things in perpetuity. You can either hand them off to the client. Or right. If you're on a team, you can just run these things for a few hours and essentially have images for an entire quarter, right? So that's the power of doing something like this. So I wanted to get into style ref. That's what I wanted. So

Isar:

let's get into the example of your one template. No, go back to the one that let's just give an example of what a template may look like, literally just read what it says. So people understand what a template of a prompt can look like.

Drew:

Yeah, these are very simple, but I did take these from. the ones that I did for the visual examples that we looked at earlier. thinking about, first who my subject is, right? So what's their role? And then what, what about them, right? Age, ethnicity. Then I'm thinking attire. Okay. Thinking about the action that they're taking. Okay. and the background that's in place. lighting isn't so much important anymore. you, it could be, but from a style reference, this is pre style reference. When I put this one together, essentially you can capture that same type of light with a style reference, right? So that maybe isn't super important. And then, if there are certain I would just say like facial expressions of the person or words or tokens or phrases that describe the scene that you're trying to create. is it a, is it a whimsical thing? Is it a moody thing? And some of that could still go into style reference as well. Probably not super important. So I would just say, focus on. Really the subject, the action, the location. and if you start there, you can build in some of those other things like facial expressions or camera angles, just to create an entire scene of what you're working with. Okay. So these are, just something that you can play with and maybe experiment with this with your own brand. If you want to be hands on and actionable with it, this is a good starting place. I'm going to bump out of here because, now we're going to switch over to more or less how you would apply. the style reference. And let me see if I can find a particular example here. this is while you're doing

Isar:

this, those of you who are watching the screen at the end of the image. There's an optional section that says, bring in a reference image. So an external image that you have, either from your brand or a third party or whatever, so you can upload images and then use a parameter. That's called IW. So those of you don't know Midjourney has a lot of parameters that you can add. You add all of them in the same way. You add, Sorry, hyphen. And then it's either two or three or four, letters. So in this case, it's IW stands for image weight, which means you can decide how much you really want to lean on that image as the anchor to the new image that you're creating. And then. And the numbers, if you're using the Midjourney website, it's between zero and three, if using the old discord, it's between zero and two, the higher the number, the more it's going to use the image that you uploaded as a strong baseline for the new image. And so if you have a thing that you're like, let's use your example. Instead of trying to explain in 50 different ways,

Drew:

I think this is going to be a double whammy here because this is both use of style reference and consistent character.

Isar:

So I'm saying, let's think about the metal claw that you're talking about, instead of just trying to explain it in five sentences, which then you lose Midjourney because it's too much. You can just bring two or three images in one image. Right off that claw. That's so it's one from this and that, put them into actually one JPEG and then use that as an image weight. It will now understand what is this thing that you're talking about. And you can start with a much more solid starting point than just text.

Drew:

100%. And so I'll use that particular strategy in situations like that, right? There's not a lot of information in the database. I've gone through different synonyms or names of things, right? So like using the same company, Excavators. We're actually six months ago, hard to create in Midjourney because it didn't really have enough learning data on them. there just wasn't a lot now, no problem, but the claw, for example, is still hard. So in that case, I need to supply Midjourney with something. It does not have and use something like an image way to drive closer to it. And so without getting too in the weeds, you're describing What you want in the prompt, but you're also taking advantage and using for leverage this image to get you closer to where you want to be. Awesome. So that's one great usage. And let's break these down maybe into three things, right? We've got image reference, we've got style reference, and we've got character reference. Yes. Accomplished different things. Sometimes you use just one, but sometimes you can use more than one. And there are advantages to each, but I'm going to show you an example. I think of one that incorporates style also incorporates persona. And let me just back up and explain what we're trying to do here. So my client says, Hey, look, we've got somebody that is, what's called a trader. in this business and they're not necessarily in the metal recycling yard, but their job is to entertain, do business. They meet with clients, they're on the road. So they spend a lot of times in airports, hotels, at dinner with clients. And then, let's talk about who specifically we want. And so what we did was we identified three different personas and then attached. like one particular character to each persona to try to build out this same person in multiple scenes, different locations for camera angles, et cetera. Perfect. Okay. So now what we have on the screen, is we're showing a picture of this character. We're showing a picture, that was using a style reference. And we're also showing you the prompt here, right? So this prompt is Instagram photo, a polished 40 year old Pakistani businessman. He's wearing sunglasses with a button up shirt tucked into chinos. He's looking at his phone at an ethnic restaurant, candidly smiling. The background is light. And so there are a couple of things in here that are very specific to the brand, the age, the ethnicity, what he's, what he's wearing, where he's at, and even maybe the expression to some degree too. So all those things can be interchanged. But what I did here were two things. One, I used a style reference. And so backing up for a second, this company's colors are green and orange. Most of what they're wearing is orange because they're in the yard, right? So they're wearing like high vis vest. that wear an orange hard hats if they work in the yard, etcetera. And so there's like this warm lighting that comes into play. And so I took this image clicked, copy image address if you're using maybe an older school approach and then dropping that in the prompt. dash S ref paste, right? So that's one way you can do it. The other way you can do it.

Isar:

So again, just to slow down for people who don't understand what this works again, the, at the end of each prompt in Midjourney, you can add these parameters. One of those parameters is style reference, which is S ref that's it's dash space. SREF, and then you drop in an image, but the way you drop in an image is you actually copy the link to the image that you want by right clicking on the original image and then dropping that at the, after you put the SREF, parameter in there and you can actually use more than one, right? You can use a bunch of images. Correct. SREFs.

Drew:

Correct. there's simpler ways to do it. Now, discord, with image weight and style reference and consistent character, like it became a lot of URL strings. Now with the new site, you could just take this image. And let's pretend this was the image I wanted to use for style ref. I can just drop it in my prompt box. And then you get these three options here where I can select either. I'm going to use it as an image reference, a style reference or a character reference.

Isar:

So again, to explain what's on the screen, when you're looking at an existing image already created in Midjourney, you can drag, literally grab it and drag it into the prompt line. And then on the new image, it shows up as a little tiny, like thumbnail in your prompt. It shows three little buttons. One looks like a face. One looks like an image and one looks like, I don't remember what the style paper clips. Like a paper clip that is like the style ref. So depending on which of these buttons you press, Click it knows what you want to use the image as you want to use it as a reference to the overall image. Do you want to use it as a reference to the style, meaning the lighting, the colors, the atmosphere and all of that, or do you want to use it as a character? Meaning in this case, this Pakistani guy with the beard and the sunglasses and buttoned up shirt. So you can pick whichever one you want and you can actually upload the image twice and pick two different things and it's going to do both.

Drew:

That's right. and this goes back to you can use one of these things, or you can use multiple. Things right. And here's why you're here's just a, an example of using multiple things, right? I'm not only trying to transfer the style over in terms of that look in the field, but I'm trying to get the consistent character, right? And so the best way to do that is to use character reference, not image reference, because I'm not necessarily trying to recreate the image. I'm trying to recreate the person, right? so here I'm using style and consistent character, but yeah, it's really easy. Now, all you have to do is drag these things in. You can select them right here. if I wanted to select both or use this twice, now I've got two options here on the screen that I can select and multi select. So say all that to say, I'm trying to do a lot of different things with this image, right? And so once I've taken, an image that I like, I use that as style, then it really comes down to, how I want to, how much of that style I want to apply. And so this is the second part of image reference, style reference, and consistent character reference, which is how much weight do you want to establish with that feature? So there are use cases for using. that scale all the way at the top, all the way to the bottom middle. And so these are things that putting in the reps or trying different things or experimenting, you're going to have a better indication as to maybe which direction you should go. The more that you play with it, but you have to play with it regardless, right? Because something may wash out the character. maybe he's not wearing exactly what I want. Maybe he doesn't look quite the way I thought he would consistently. Maybe the style overweights the prompt. So all those different scenarios come into mind. So that's why it's important to really just play around with different values to see what you get in return,

Isar:

but let's talk about the actual practical values. How do you put them in and what are the limitations? what are the, what is the range of each and every one of them?

Drew:

Yeah. Okay. so they're all different, right? Let's start with, let's start with image weight. So image weight and all of these are used as parameters, like as our set. So these would come at the end of the prompt. if you're typing them out or if you're just dragging them in, that's fine too. But, image weight, like if you're typing it is I, dash IW. And then this value is between really zero in three, zero, meaning you don't even need to put it in. it's not even, it

Isar:

doesn't make any sense to put a zero. Just don't use the image.

Drew:

Just think you wouldn't use it. And then three is, trying as much as possible to replicate that exact image. Although that's not super possible with this technology. And it's not very consistent either because I've seen times where it's I get a better output or result using an image weight of two or a 1. 5, which is right in the middle. Then I would like a three, for example, which should give me the most, the most identical or similar, so any of these things require, testing and playing around because of the consistency issue, but they're super powerful, right? image weight between zero and three, you can use decimals, right? So anything like a one point, you could go from like a 1. 25 or you can do a 2. 75, or 2. 75, anything in between there and play around with those things. And I think if you're going for speed and you really want to be efficient with this process, that's where permutations come in handy, which is using, a curly bracket. And maybe I want to try an image weight of one. I want to try an image weight of two, and I want to try an image weight of 2. 75. And then I would just, basically take my prompt and then it's going to treat those as three different jobs. It's going to use one.

Isar:

Pause you for just one second. So you taught me that in a one on one session, like six months ago, and it changed my life when it comes to Midjourney, because these parameters change a lot in the image. And really the way I used to do it as it was just to run it as okay, image weight one and run it and see how it goes and then run it again at image weight 1. 5 and see how it goes. And then run it. And that's not very efficient. now the reality is you can use after the parameter, instead of just after the IW in this particular case, the image way, but it could be the same thing for SRF or any other parameter. Instead of putting one number, you can use the squiggly brackets and then put value number one, comma, value number two, comma, and put as many as you want. then it's going to run it several times and it's going to show you all the results. So your starting point is four variations of the same image with different image weights. Oh, I really like that one. And like Drew said, it doesn't really matter. What was the number? Was one the best 2. 5? It doesn't matter. This is the one you like the best. You just have several different options. And now you can use that image. As your first reference to the next version, which now narrows down to, okay, this is what I'm looking for.

Drew:

This to me is one of the game changing. And I hate that word. I just don't have a better word for game changer. it really is one of the best features because one, it allows you to basically speed up your whole iteration process. Two, you get to play with all these different values, right? Because. I don't want to think about what if I didn't type this number in and I maybe missed a great image, right? Like I want to explore all the pathways I can as quickly as possible. Yeah. so it really does come down to these micro decisions in a tool like this, let's just start with one image. How many decisions can you make just with one image? you could You know, like one prompt. you could rerun it. You could subtle very one image. You could subtle strong an image. You could, in paint with a very region where you want to change something specific within an image. You can zoom right. And so you think about that option with just one grid. And it's yeah, there's a lot of options there, but then you can start adding in permutations with different weights or different styles or different stylized numbers even, and then you're exploring more of the breadth of what's possible with your prompt. And you're not leaving anything out to chance, right? You've just explored more doors that could be open. And to me, it's all about the experimentation and then taking just a brief second to say, what did I learn from that experiment, right? Did I learn that higher stylized value worked better with this longer prompt? Or did I learn that, a lower weight or whatever it is. I'm trying to analyze, what I just, experienced in that same token. So I think that's super, super important if you, especially if you're creating images outside of just personally, right? So if you're doing anything for a client or, anything for your business, maybe you're a marketing leader, for example, you have to utilize something like this. for time efficiency. Awesome.

Isar:

I want to touch, I want to summarize quickly. A lot of things we talked about because I know we, we dove into a lot of tactical things that people might be lost. Yeah. So I want to recap that. And then I will start with something else that you mentioned that is very important that I've learned as maybe the biggest trick there is. As far as getting consistent images. So I will start with that. And then I'll do the recap because otherwise I'll forget you talked about in painting. So one of the capabilities of all of these tools, by the way, is in painting, which means you can pick a brush, which is a circle that you can make really small or really big. So you can do it really fine or really broad and paint a section of the image. And then prompt just that section of the image. It is the only way to get exactly the same rest of the image as the original image. So let's say you got your character or your product or your street or whatever the thing that you got, He's Oh my God, this is perfect. I just want to change this one thing. If you'll try to reprompt it with whatever. You put the highest image reference and the highest style reference and the highest character reference. It will never be the same image. It will never be. However, if you use the in painting section saying, okay, color just staying and changed as that, the rest stays the same. And so if you have one that you really you just want to change stuff in the background or whatever, that is your way to go instead of trying, Oh, I'm going to try all these other permutations, but so this is just a tip, but now let's go into that.

Drew:

That is just one

Isar:

second, the summary, and then I'm sure you have a gazillion examples of this, so I'll let you show it, but quick summary. Using Meet Journey to create images is very powerful. The biggest problem is that it's not consistent. The way to get consistency is to use one of three tools that, or now four that we talked about, including in painting, that you can use. One is to use an image reference, which is basically using an existing image that you have that doesn't have to be from Meet Journey. Could be any image that you have, that you took, or from the internet, or whatever that you have access to. All right. You can upload that and use the IW, which is image weight value, to decide how much. Wait to put to it when it's creating this. The second one is to style reference. And it really is just trying to capture the style, the color palette, whether it's moody, whether it's bright, whether it's funny, whether it's like trying to capture the style of the image. And the third one is character. If you were able, if you brought an image from the outside, if you're able to create a person that is exactly the kind of target audience you're trying to target young, old, female, male, black, white, with a hat, with a tuxedo, with a hoodie, like if you are able to capture that, You can use the character reference value in order to continue having the same person in multiple angles, multiple shots, multiple images. this is to put everything together. Now I'll let you show some example because I know that hit a point.

Drew:

Yeah, no, this is, you're really doing a great job recapping this. So what I was going to show you too, is I think that's a great point with the very, the in painting or the very region feature, as they call it, Is like this, like one of the ideas that my client came up with is Hey, these people wear multiple hats, quote unquote on the job, So they have their job, but they're also taking clients out. they're also acting as like a. a doctor on site or a scientist or an accountant, right? Because of all these situations that occur in their day. And so one of the things that we did was, utilize an in paint tool to basically use the same character, just changing his clothes. Background. So to your point, I would never be able to create this same guy. This was before consistent character, especially there was no way to recreate this. and especially you can't do that now with the same background is, is I think, even though you have a consistent character, your background's probably going to be different unless you're using just like a plain studio, for example, but, This is super cool because I could just change these outfits. So here's

Isar:

those who you're not watching. We're looking at a guy standing in front of a slightly blurry, like like shell shallow depth of field. and in the background, it's like a metal dumpster behind him, but then going between the images, it's the same exact guy, same exact everything, but he's wearing a rain jacket. And then he's wearing a polo shirt and then he's wearing a hard hat. And like a safety jacket, but it's other than that, it's the same exact image.

Drew:

Correct. Yeah.

Isar:

So I'll give you, a cool, funny example about me when I've used it. So I'm a part of this venture capital and we're upgrading the website and we needed like photographs of ourselves. And I'm not the guy that goes and does a professional photo shoot. So I don't have a professional photo shoot. So I took a snapshot out of one of my podcasts, just a screenshot for, because I have a gazillion images of me and I, when I have a fine, a good smile on my face and I took that and I send it to the guy and he's This is a venture capital. So wearing a t shirt may not be the outfit you want to be wearing here. So can you take another image with you wearing a buttoned up shirt? I'm like, sure. And literally what I did is what you did here. I went, I opened Midjourney. I uploaded the image I used in painting to change the shirt. I put, stylish buttoned up shirt, blah, blah, blah, blah, blah. And I had another image, the same exact image of me just with a buttoned up shirt. So you can do, and the cool thing, it's not like a Photoshop. I'm trying to crop an existing shirt onto the body. It literally adjusts the shirts to my body, including the slight wrinkles. Like it looks very realistic. And those of you who are watching this, you know what I'm talking about from the images that Drew is showing.

Drew:

Yeah. so that's, I think very original that in painting capabilities. Probably one of the more fascinating uses of AI as well. so you've obviously got generative fill in Photoshop. You've got, very region and Midjourney, but it's great because you and I were just talking about, Hey, like there still are some limitations to things like, using AI for visuals, like every once in a while, you're still going to get a sixth finger or a third leg, right? But maybe you'll up, maybe you love the rest of the image, which happens to me a lot. I could reprompt, but I'm going to get a different, a different person. They're going to look slightly different, maybe a slightly different outfit, depending on what the prompt is, or right, I could use that feature and just change a certain section of the, the image. these are all upscale, so I don't know if I can do that on here, but like an example would be clicking very region, and I'm just showcasing this for anybody visually watching how you would make that selection. So you've got two tools that you can use here. Let's just say if I just wanted to change what's above his pants and I wanted to keep the pants, then I'm just going to account for that. And I don't have to be super, super specific with this because the technology is pretty good, but I'm just going to outline the part that I want to change.

Isar:

So to those of you not watching, he's basically using like a lasso tool. So it's like holding the button on your mouse and dragging roughly the line around in this particular case, the jacket that the guy's wearing.

Drew:

Yep. And also wearing a hard hat. So I'm just selecting that, maybe I want to change that. And then I would just edit the prompt, so this says, he's wearing an orange safety vest with full length, Navy blue overalls underneath. So if I just wanted wearing a, let's just keep it simple, like a white. t shirt, right? I would just hit submit and then I'm going to get something back. and then I've kept the integrity of the rest of the image. Yep. And so that's obviously important with even like the finger or leg example, right? I could just select that part that I want to remove. And in that case, I don't even need to type anything for the prompt. I just submit it. And the technology is good enough where most of the time it's going to correct it for you. So I think that's fantastic tool to use as well. Cool. As you get closer to your final image and is are you touched on this but the process starts very wide. Experiment, experiment. And then you're getting something where it's like, Oh, there's something to this. So maybe I'll very subtle it or very strong, or when I go down a different path, now I've got something that feels like much closer and I just don't want this in the image, or I want to change that. now you get down to maybe a very region and they're like, okay, this looks pretty good upscale. And if there's maybe something else like super minor, you want to change or. Adjust. Yeah, you could take it into Photoshop, but I'm trying to do as much as I can in the tool if possible. And so that's why, knowing all the features, the capabilities, some of the things that we talked about today, but that's how I know, the different things that I can do is just knowing how all the features and capabilities work. So if you're exploring this tool, I can't stress that enough. know all the things that, are possible that you can use to your advantage in order to troubleshoot some of the problems that you might run into.

Isar:

Awesome. I want to say something quick about what you said and then have one more question and then I'll let you go. I know you have other stuff to do other than just talking to me, even though I know you and I can talk about this. You and I can geek about this for probably hours. But, the one thing that I want to add that has to do with what you said. Remember, you're not limited to just this tool. Like you, you can say, you know what, I can fix this small thing in, in, in your case, Photoshop, but this could be Canva in five seconds. Okay.

Drew:

Yes,

Isar:

don't try for an hour and a half to figure out how to do it in Midjourney. Take the image, bring it to Photoshop or Canva or PowerPoint and add whatever change, whatever you want to add and be done with it. there's, it's perfect. I do that all the time. Like when I want to add like an overlay, something on Midjourney, it's almost impossible to do. Like my logo or text or whatever, I take the image, I bring it to Canva. I add whatever I want to add and I'm done. And the, that extra step takes me five seconds and would have taken me two hours and probably not be as accurate as it is in Canva. So it's perfectly fine to mix and match multiple tools in your process. Once you figure this out. The other thing that I will say, and then I promise you the final question is that this sounds like A lot of work and experimentation in time, and it is like it is a lot about trying things out to see what's working, what's not working, what's getting you closer. Once you get closer, he's starting to dive deeper into that rabbit hole of that initial image, as Drew said, but compare that to the alternative. How would you do this? Before this existed is usually hiring a crew of photographers and actors in the site and lighting. And it's, it would have been not an hour of experimentation, but three days out in the field, that's going to cost you between 20 to 40, 000. And so while this sounds like, Oh my God, this sounds like a lot of work. First of all, as you get better, you'll be able to do this with a lot less steps and attempts, because going back again, I loved it. What you said, Learn from it. what did you learn from this little experiment? So next time you don't have to do 70 different options. You can do seven because you know which path to go already. So think about that. It's always a better option than the other alternative deal we had before now to the question. I know there's a way to save a style. So instead of, if I want to use the style again, now I'm like, okay, I got to load the image again and use S ref again, like I know there's a way to save a number. of a style. And I would love to know how to do that. or that's still a discord thing and it doesn't exist on the website anymore.

Drew:

no. so there are two, I'm just trying to think of the best way to explain this to maybe somebody that's not super familiar. So it doesn't get too diluted or confusing. So essentially what you're talking about are style codes, correct? Okay. So there are two different ways that you can create sort of styles. You've got one using an existing image, preferably an Midjourney image, which is what I did in that example that you saw, or right there are what's called style codes in this feature called style random, right? and I'll actually run it for where you guys here while we're looking at it, for those of you following along with screen, but I could just type in, let's just type in girl and I'm going to put S ref so dash S ref. And you just type in random. Okay. And then this is going to give me a random style code that then I can use. And to your point, just copy and paste. Now you can't, there's not a great place to save that currently, other than just like keeping a little database together, which I've created because I had to, but, you can apply that same number over and over again. And essentially what this has done is MidJourney is taking all these different styles that it's got available in its latent space. And it's making them accessible for you to use. And I do this as an option, because I found that when you use a style reference photo, sorry, if this is confusing, but if you use a style reference photo, sometimes it takes a little bit of the image into consideration as well, like it's the subject, whereas the style codes are purely style, right? And so if you've got some style codes that you've come across. Save them, use them. Some are great at more photorealism, some are great at illustration, and I can show you a few of those as well.

Isar:

So I have an interesting question, which I don't know the answer. I'm really curious. Is there a way for me to know what's the style code of any image that I've created, or there's no such thing? if I, if you have one of those images, you can click on it. Is there a way to know what's the style code of that image? No. So the only way is to create a new image with style, with sf random, and then it gives you the code, and then you can use it.

Drew:

Yeah. so what a lot of people that, are like deep into Midjourney are doing right now is they're doing something like this, where they're typing in a short prompt, they're typing in dash S ref random. And then they're running that

Isar:

500 times.

Drew:

Yes. Because they're trying to just unlock. Styles that they might use or other people might find interesting and some of them suck, right? some of them are trash and some of them are really good. And the unfortunate part is right now, as it stands today, this may change by the time this goes out, but there's not like a dictionary.

Isar:

Yeah. Clearly trial and error.

Drew:

Yes. And

Isar:

So how do you see, where do you see the code of the images you just created?

Drew:

Okay, perfect. Great question. So then, I piped Randa, but now I've got an S ref and it gives me this number. 3, So now, if I wanted to use that style, I would just use S ref in that number. And for this example, I'll just use the same prompt, but with that number.

Isar:

So those of you who don't see, it was a very short prom, but what it created is like psychedelic image, four options, obviously, because there's always four, these psychedelic images of as if a very scary movie on LSD. A kind of style, and, but if you want to create these kind of images, now we can apply this to everything. This could be an office that looks like this could be a street that looks like this. This could be a drink that looks like this, like whatever we're going to use that code now is going to create the same feel, which is very powerful if you have the few that you actually like, because they generate the images that you want.

Drew:

yes. And so this is right us taking that number and just applying it again. And then this is where you can really get into, really the nuances with within the style code. And this is I think the next step of using it is playing with the style weight right at the beginning we were talking about how image references have a weight style references have a weight consistent character references have a weight. Yeah. So the

Isar:

code is dash SW for style weight. And then you can put in numbers between what's the range.

Drew:

This one is a zero to a thousand goes up to a thousand. I like to, so again, using permutations, I like to see where this fine line of this is too much style or not enough exists. So they're all a little bit different. So I'm going to play with a few different values here just as an experiment and kind of show you where this goes, right? So I'm typing in five different variations, stylized 35, 50, 75, 100, and 250. And then you'll see how the application is different in these. And so that's where it's that style, default, which goes to a hundred, which, what, which is what we saw, maybe that is a little too strong. So maybe I want to pull it back or it's not enough. And so this is where, I think, the fun really happens because you can dial into that exact level. That's like. taking in your prompt into consideration fully, but also just applying enough. that makes a big difference. So I'll just start for those of you that are watching for those of you that are not, we're just going to quickly look at each value just to show you the differences in them. So these are from style weight 35. So really a real low application of it. Then we go up to 50, right? So you see a little bit more of the style in there now, 75.

Isar:

Yeah, that starts getting to look very much like the original images.

Drew:

Yep. And then this is a hundred. So these are the ones that we default, defaulted to with the first one. And then here's two 50, right? these look completely crazy or abstract and psychedelic. And so that's where it's, it can be super interesting and maybe for brands, right? it's Can I find the color palette that exists for my brand and apply that with the weight that makes sense. And just before I go, and maybe what could be helpful is our, is maybe I'll share, maybe I'll share this document with the listeners because. Just send this to

Isar:

me. I'll add this to the show notes and then people will be able to have access to this. Perfect.

Drew:

Like I just grabbed like a hundred of the ones that I found in light. Okay. Put them into the sample image. Here's the code. And then just described them with tags. so that's phenomenal. So that could be helpful for some folks too.

Isar:

Drew, this was incredible. Like we covered so much. I'll do a very quick recap, but it's creating stuff that is worthy of being published as part of a marketing campaign or your website or brochure or whatever it is that you're creating requires. Knowledge on how to use these tools and Midjourney is still the best tool out there. at least for creating high end stuff. if you want to create quick and dirty, I like Dall-e very much. I use it a lot just to create quick and dirty stuff because it understands me faster. I don't have to be a prompt guru to, to get the images. But if you're looking for quality, Midjourney is still the gold standard. And we touched on, A few advanced Midjourney capabilities that will allow you to really create professional looking style, dialed in images that you can use for anything you want for your business. And this was a fascinating conversation touching on several different ways. You can use these advanced capabilities drew. This was. really amazing and insightful. You're obviously one of the best in this, at least from the people that I follow and know if people want to follow you, work with you, know more about you. What are the best ways to do that?

Drew:

Best way to find me is on LinkedIn. Probably if you're not on there, I'm also on Instagram, right? All those basic channels. So you can find me there as well. But, last thing, I guess for me that I just want to make sure people know is while some of these tools are not quite at a finished product, they are most certainly like a tool like MidJourney ready to be used today. So at this point, there's no point to waiting any further. These tools are only going to become more and more advanced in terms of their capabilities the complexities right the workflows that you can create while using these things what we talked about today. there is. this idea that you should jump in now, don't think about this in terms of the sunk cost fallacy, if it's too late, or it doesn't make sense to jump in now, or I'm going to wait till later because it's only going to get harder, I think for people that are just getting started. So it's a great time. And if you have any questions. I'm an open resource. I love talking about this stuff, obviously. so feel free to reach out and, yeah, we'll talk some more.

Isar:

Awesome. This was great. Thank you so much. again, in a few months. Yeah, let's do it. Awesome. Bye everyone. Bye.

People on this episode

Head Shot

Isar Meitis

Host