Master ChatGPT image generation to build your brand faster

Navid Moazzez: In this session we will talk about ChatGPT image generation and how you can build your brand using these tools. And my guest for this session is Mitch Asser and he is been using ChatGPT in a lot of different AI automation tools to grow his businesses and also scale his virtual summits and, other things in his business.

Now he's building BigMoves, which is an AI automation and tools company. And, he was actually one of my first students inside of Virtual Summit Mastery and he really took it to the next level automating all of this for his last business, which was Pet Summits. And yeah, we probably will share some of this, but now we're gonna dive into more of how you can really go from a beginner with ChatGPT to more an of an advanced user when it comes to image generation.

So with that being said, warm, welcome, Mitch. Thanks, vid. Excited to be here. It's gonna be a great session. Yeah, let's talk a little bit more about this. Maybe you can share a little bit how you got into it, and then we can, uh, dive in. I know you have a big screen share, like a lot of stuff you wanna get into.

And I saw the notes before this session, so it's some really exciting stuff. Uh, but yeah. Anything you'd like to share before we kick this off?

Mitch Asser: Yeah, absolutely. I wanna make sure that this session is as valuable as possible. So we've been, um, building an agency Make Big Moves, AI, where we are working with clients on a day-to-day basis, building all types of different workflows.

And what I noticed is that the capabilities required in order to automate workflows using AI are very different in terms of what is possible inside of ChatGPT inside of the dashboard when we're generating images or analyzing images. So in this presentation, I'm gonna step through how you can go from a beginner, like using the ChatGPT dashboard and what some beginner prompts would look like, and then how you can use different tools, whether it's GPTs, um, different ways of prompting the AI in order to create much more advanced, uh, let's call 'em templates so that they become scalable and allow you to automate if you do want to go down that path.

Because ultimately there's, um, a, a whole different. Uh, there, there are many different ways that people are using ChatGPT, and I find that 95% of people are actually using it in the slowest way possible, which is, you know, you do one job at a time in the dashboard and then you move it to your other tools.

But we can actually do it so much faster using automation. So, but yeah, we're gonna go from beginner to intermediate to advance and just go through that pathway, make it as educational as possible, and interactive as possible so that, uh, yeah, you can be much more streamlined in, in your approach to generating high quality images at scale.

Navid Moazzez: Yeah, so this is not just your average ChatGPT session. This is gonna be some more advanced stuff as well. So you can take away and hopefully implement this for your brand, your business, and, you know, scale things because you can do so much with this. And that's not just with ChatGPT, bt you can use it together with other tools.

And I know you are doing this, Mitch, you have some crazy automations and stuff like that with n it n and I'm not sure if you'll dive into some of this as well, and maybe in another session we have as well. So you're gonna talk more about the automation side of things, but how about we just dive in and then if I have some questions along the way, I'll just, uh, chime in here.

Mitch Asser: Yeah, of course. Sounds good. So let's bring up the screen share and we can start with the, the Google doc here, uh, as well. Are you, is this, this up on the screen now? The, the Google

Navid Moazzez: doc? Everything is up, so you can, you can just, uh, start here.

Mitch Asser: Perfect. So I mean, there's, there's gonna be, uh, multiple things that I take you through.

And the, the way that I'm gonna do this is just working through the dashboard and, and, um, maybe creating a few examples. We are going to improve some prompts, but also I'm gonna tease in a few of the automations that we've built for clients so that you get an understanding of like how far you can take this when you're actually automating it in the API.

And if any of those words are scaring you, then don't worry. It's not as complicated as it sounds. But we are going to dive into all of this. And then for a more detailed overview of automation and how you can automate so many different types of workflows, we're gonna cover that in another session during the Summit as well.

So let's dive in. Uh, I have some notes here. We can make these notes downloadable as well. Um, but basically we're just gonna start with like, what a. Typical beginner would do when it comes to generating images. And the goal is simply to show what a simple prompt looks like, um, what beginners would probably start with, and then, uh, think of like the, the key elements that would be needed.

So something that I've become absolutely obsessed with this year is playing paddle. So if anyone who doesn't know paddle, it's basically a mix between squash and tennis. Um, played on a smaller court. You've got the walls around you that you can use. So much fun. But we're gonna use this as our reference.

So I'm just gonna paste what I would call a beginner prompt into, uh, ChatGPT. We might even make this, uh, a, let's, let's do like a wide image to make sure it's easy for everybody to see. And we're just gonna ask it simply to create this image of a paddle game with a modern paddle court, but set in ancient Greece times with an ancient theme and ancient Greek players.

The reason why I did this is actually, um, a funny story. I was, I started a paddle, uh, Instagram page called Up for Paddle and w it was, it was me building a, uh, video meme workflow just with, uh, make.com and Airtable. And I was posting like a few memes per week and had a few of those go viral. And I just came up with this idea once to, um, do like a funny.

Wisdom quote by a guy called plos. So it's kind of taking that theme of Socrates from ancient Greece and then, but giving him the name of plos and it went viral. Like I think we had, uh, I don't even know, like hundreds of thousands of views on that very quickly. Um, so that's kind of been the, the, the reoccurring theme in some of my test, uh, images and different, uh, different videos that I'm creating with AI recently.

So it does take a little while to create this image, which is why I'm blabbing along here. But, uh, ultimately it's gonna create an image that I would say most beginners would, would get, and the image will probably be impressive. Um, there occasionally. It's getting so much better now there, but there's occasionally smaller mistakes and however, it's not that detailed.

Right. So let's give it a, a moment here. Maybe we can edit this bit out of the, of the talk.

Navid Moazzez: Yeah. I mean that's, that's always when you're doing something live, it's always taking longer time. You know, that's, that's kind of a thing. But, uh, yeah, we'll see. Actually it's interesting. I, I have the same experience as you, like when you start, you just have some basic prompt.

But what I found also with, you know, increasing my productivities, you actually, I, I speak a lot to ai. I use like super wis. I dunno if you've tried that, man. I use super whisper, whisper flow. I never speak to any tool. It's amazing because I, 'cause when you are describing an image, you basically need to be, give vivid detail and to write, that takes a little bit of time.

Yeah. But I'm sure you get into some, some stuff, how you do it as well.

Mitch Asser: Yeah, yeah, absolutely. We're definitely gonna dive into that. So yeah, it's created a decent image, but it's not, I'm not really sure. That's what I'm looking for first, um, is that the two partners are looking at each other instead of their opponents, and then there's a ball in the middle that's going to just go to nobody.

So, not ideal, but it's at least a starting point, right? The the next, the, so when I went from being in this stage and like wondering how to create prompts, I was, I was actually in Midjourney and I could see these really long prompts getting created and they had the same types of themes over and over again.

Even though there were different users creating those, those prompts, I was like, well, there has to be some sort of SOP some sort of documentation in order how of how to prompt correctly. Now, most of these platforms actually do have really great documentation telling users how to actually use the platforms effectively.

Most people don't read them. So of course, like humans are lazy in nature. So of course we just try and do it the the fastest possible way. But there are also GPTs that have been created. So one of my favorite ones, which I use not just for Midjourney, but for all of the, the models are this mj, prompt Generator V six.

Not sure how this became one of my favorite. They have seven now, by

Navid Moazzez: the way. They have the VI dunno if it's the same. Oh, there, I think it might be the same company, but yeah, because the V seven, I think it, no, I think they're different. Slightly different. Yeah, they're different. Yeah. Maybe. Maybe different.

But yeah, I use, I use Midjourney because there's a lot of like creative people. It's a different user base, a little bit, like I feel like the photography type people are more like the ones, uh, that, that would use something like Midjourney. But I, I think it's the best results you can get. So it's like you can get really good results with chat GBT.

For sure, you can, but like, if you want a specific style, I think a lot of people go to Midjourney and things like that. So.

Mitch Asser: Yeah. Yeah, I agree with that for sure. So I'm just gonna use this while and get it to, to run while I talk through some of the others, but we're just gonna ask it to improve this prompt.

And it's gonna give us a few versions, but also there's, there's GPTs built for all of, basically all of the models. Yeah. So there's one here for Flux if Flux is your favorite. There's another one here for Ideogram if Ideogram is your favorite. Um, there's probably some for Imagine and, and all of the others that are out there as well.

Yeah. So now it's gone through and it's provided us with a few variations of that prompt, but in much more detail. So let's just grab this first one here and we'll come back to the original screen. And we're just gonna look through this prompt. So it's got, it's built in the. Um, with the extra parameters required for Midjourney.

Like if you're in mid Midjourney, then you're gonna write Imagine in order to start creating your prompt. It's for the Discord. Now, the end, actually they

Navid Moazzez: have the web app, by the way. I just know, I just played around with it. So it's actually ChatGPT. BT is a little bit delayed with how they're thinking about this.

So if you're using the web app, you don't need this. It's in Discord. For example, the 16, like some of the chaos stuff and you can still add these parameters, is is fine, but you don't need, for example, to add 16 by nine because you can already set the aspect ratio inside of the web app. I would recommend the web app actually, it's really good.

I would not recommend the Discord too much anymore. Personally, at least I haven't

Mitch Asser: used it since, since the web app came along actually. So, but yeah, the typically what you would do inside of Discord was add the aspect ratio and then the version that you wanted to use. Um, but ultimately, let's get this one generating while we speak through it, the, it's now created.

Much more dynamic prompt and much more detailed prompt. So it's saying cinematic photograph, it's saying dynamic action, mid game players dressed in tunings and sandals. Um, the chord is modern with glass. So this alone takes your prompt from like maybe like a two or three out of a 10 to maybe a seven or eight out of 10 immediately just by doing these types of things.

So when I first came across this, I was like, okay, we, if we are going to, um, use these things with our clients, how do we then structure, or how do we then understand what these types of variables are? So I would actually just open up a new thread and I started chatting with, uh, chap GPT to reverse engineer what was included in this GPT, like the, the backend of how this GPT is built and what we now do.

So we, we've got like a whole bunch of principles, which maybe I can talk through some more. Um, inside of Miro board here, and we can talk through some more in the, in the next session. Yeah. But basically we, we can, we broke it down into the seven elements of the perfect AI image prompt. And those elements are the physical traits.

So, um, in the example we had physical traits of, um, where was it? Like players dressed in traditional ancient Greek jus and sandals wearing model, paddle rack, uh, with model modern paddle rackets. Um, we have the pose. We have, uh, the expression, we have the background, so what setting is it in? We have the lighting.

So lighting is especially effective to change if you're doing like portrait images. So if you're trying to generate an image of yourself, it could be like studio lighting or natural lighting. Definitely changes things up a lot. Um, details. So like any fine textures, like skin texture, sharp reflections in the eyes, uh, and then equipment.

So this was huge. When, uh, when we started adding in equipment each time, so like, and actually giving real models of real types of cameras and, uh, lens sizes in order to, to get the, the, the right image or the right, uh, um, type of shoot that we're looking for, the right type of shot. So like these seven elements, as soon as we started adding these into every AI image prompt, that's when we just elevated the quality that we were getting from, from each of our requests.

So let's just come back now and see if we've got this image. So it's still. Not entirely accurate. Like it's, um, I, I mean, paddle is actually quite a hard subject. It's quite new. Um, not a lot of data around paddle for a, the AI to, to learn from. So it's quite hard to generate images, but it's a huge improvement compared to that first one, right?

Like the, just the, um, the, it's just a more crisp image. Um, and yeah, I think it, it works a whole lot better. So that's how we, we went from like beginner to intermediate. And I think for the next one, I'm actually just gonna open up some thread that I was just playing around with yesterday, rather than generating them, um, from scratch.

But you can take to a whole new nother level when you start to use, uh, what's called JSON. So JSON is, you know, part of what developers use in order to, um, develop certain workflows. And we're just going to go, let's go back up to, let's maybe start with this one. So this was the same, uh, work type of workflow I went through yesterday while preparing for this, which is, you know, the basic prompt that we wanted to create, which is the futuristic AI utopia.

Then we improved it with the Midjourney, uh, prompt Improver created this one, but then we wanted to improve it again, and we started creating some JSON prompts. So, JSON is simply creating variables like the main prompt, like the lighting, like the atmosphere, the style aspect, ratio, the render. Store the style is in there.

Again, materials, lens resolution theme, um, robot design, because we're talking about robots, lighting effects. And then we were able to get a much more detailed, uh, image from doing that. So the question then becomes like, how do we build this JSON? And there, there's not really any right or wrong way. Um, so you could take these seven elements and ask it to build a JSON script with these seven elements, and then you create the prompt that way.

Or you can simply just analyze existing images and ask it to build a json script based on existing images. So that's what we did here. Just coming

Navid Moazzez: in here. One thing I think it's really important for people to understand this, you don't actually need to be a prompt master. As you, as you see here, you can either write.

To have a conversation with the LLM of choice. I, meaning ChatGPT is usually pretty good with prompts and stuff like that, so that's why we use it here. You can use Claude as well. Claude is pretty decent as uh, too, but I think ChatGPT is just more a little more updated with certain things, uh, when it comes to some of this.

And, uh, yeah, then you don't need to, you don't need to know code and you know, this JSON and stuff like that, I don't know it, but I can make a really good JSON prompt for Veo 3, for example, to create videos and I don't know anything about that. And I can make something incredibly cinematic that I would never be able to do without, uh, you know, all basically having access to these, uh, you know, AI to get a better, better output essentially.

Mitch Asser: Absolutely. And if there's anything that you're ever unsure of, then you just ask ChatGPT what it is. So you could say, what is JSON? Help me understand that better. Yes. And then it's going to tell you what that is, and then you're like, okay, now I know how to use this JSOA little bit better than I did, you know, 10 seconds ago.

So yeah, it's, uh, it's such a great tool in order to, uh, just improve your capabilities in terms of what you wanna build, in terms of becoming a better creator, a better business owner. So, yeah, it's really, it's such, so, so awesome. But let's say that you have a, an image style that you've seen somewhere else and somebody has created it.

You can go into Sawa, so as part of chat, GPT you on, I dunno if you can see on the left there, but you'll see it on your left hand menu. Um, SAWA. You can go into there and you can generate images in there. You can find images that others have generated. So like it was a cool, uh, SpongeBob Squarepants one that I found, and you can just download them and then ask ChatGPT to build a json based on this style.

And I actually have that, I'll find that thread as well. We'll go through that one. But another one that we have here is we create these, uh, posts on Instagram. So I thought, okay, let's see if we can get chat JPT to generate something like this. 'cause currently we just do it manually inside of Canva, um, and see how good it is.

So it generated, this JSON here didn't actually get it right the first time. So it does help having a little bit of an understanding of what all of this means. And what I mean by that is I asked it to generate, like, my, my instructions were very broad in terms of, um, what I asked it to do. And it generated this template here and it says background, and then it says image URL.

Background image, URL. So I actually didn't want to add in a background image. URLI wanted the AI to generate the background image itself. Um, so we had to change it, uh, which we did down here. Remove all instances of logos, add the background image. And the background image will not be a URL, it should be generated by ai.

So we simply just asking it to correct itself. Uh, so then it, it has another go at it. And then I said, okay, that's good. And then we posted in a news announcement about notebook lm. Um, we asked it to create a new JS Om using the same style as this image, but changing the actual prompt, like the Descript, the background Descript, and the text adding in the text.

And then it comes down, oh, what happened in, in this section? It just section here. Could

Navid Moazzez: you get the, basically how I understand Jason prompts, because I used it quite a bit for Google VO three because I saw some really, really incredible prompts for this video platform to kind of get the best result. And I, I understood it was like, it makes it more structured, like you have the, like the, the details very structured than, than if you have like long text.

But I've seen like two approaches to this. Basically, I saw a guy who's charging like 50,000 to a hundred thousand. He said it was kind of a little bit overrated, but at the same time depends I guess what you, what your use case is. 'cause uh, he basically had a, still a structured prompt in text, but maybe it's more difficult to kind of get it right.

He was, I guess, more advanced, advanced with storytelling and filmmaking so he could maybe get kind of the text prompt to be right. But maybe for someone else it's easier to just use. Jason prompts. I don't know if you have tested kind of both methods. Yeah, uh, yeah,

Mitch Asser: uh, I mean, there's, there's no right or wrong way.

So remember like if we go back to the premise of this talk, it's going from beginner to intermediate to advanced. So the reason why I'm, I'm explaining this here is because I'm gonna tie that in in just a moment to how we then use it inside of automation tools. Love it. Because once you start going into the API, you actually, it is a lot easier to build with smaller parameters than it is to, um, have a much longer No, I

Navid Moazzez: get it.

Because you are doing this for like Airtable and stuff. I know, I understand this. 'cause I haven't, I haven't, I've used a lot in a 10 and stuff, but I haven't, you you are like a master with Airtable stuff, man. So I mean, I, I leave it that up to you, like that's, that's not my 40th exactly yet. But, uh, I, I love NN and, and, uh, the CCPs and stuff like that with Claude and some connectors with chat.

Bt, but you have done a lot more with automation stuff.

Mitch Asser: Yeah, yeah, yeah, absolutely. So it, it'll all tie together in, in just a moment as we get through it. But as you can see, using that JSON prompt, it's come out very, I mean, pretty decent compared to what was, um, uh, what we started off with was just a few basic prompts.

So like, it's, it's almost copied what we've put in up here. It's missed out on the logo, it's missed out on the swipe. Those things are kind of hard to to add, but overall it's done a pretty great job in order to produce the result that we were looking for, for this, uh, type of workflow. So, um, that's one example.

Lemme see if I can quickly find maybe another one.

Uh, what did I do here? So now here I was just going through and asking it to, um, create certain styles. So yeah, that's what I was, what I was doing in this one. So what you can do is, let's say you have a certain style that you, that you really like. And let's just go down to like this SpongeBob style, which is quite different.

So what you can do is create a bunch of JSON templates where 90% of the, the JSO templates remain the same. So you might wanna keep the lighting the same, the atmosphere, the style, but it's just the main part of the prompt that you change. So, um, in this one, it. Created the prompt of like the white Mercedes.

Basically it just explained what's in this, this, this section here. I was like, okay, well let's keep everything the same except change the Mercedes to a Black Range Rover and then put them in an alley with graffiti. So it went ahead and changed that. And where did I paste the, the prompt? Um.

And then this is the image that it generated. So pretty close to the same style, right? Like this, it's got the same feel, but it's changed the, the car to a Range Rover, and then now they're in an alleyway rather than at the bottom of the sea or whatever the last one was. So you can have a whole bunch of templates set up for all of the different styles of images you wanna generate, and then you just change out the, the one part of that, JSON.

Now, where all of this ties in together is we, um, we've built a number of workflows now for clients that do these types of things. So this is one that was for a brand, and this doesn't work because it's not the, I just copied this base from our clients. But you can see here when we were creating the variables for this brand, we had all of those, um, trait or all of those, um.

Elements that I spoke about here before built out for the client so that the, the client has full control over, uh, the style that they want to use even after we've built the automation. So they can come in here at any time, and then if they want to be more specific about the physical traits of their customer avatar that they're chasing, then they can change this prompt.

If they want the pose to be different, then they can change it. So all of those seven elements across here have been added into the dashboard. And then what we do is, um, we've connected this into what I call the product list. So they've got like 700 products that they're adding into this database. And then, um, let's just open up one of the products here.

So it's like a lipstick that they, they have, and then the, we have automation set up that pulls in all of these traits from their brand. And it will send it into the AI automation in order to then generate this type of style of image for them. So you can see if we scroll through all of these, it's every single image is very much on brand with what we've asked it to do, because that, that's probably the biggest thing that I see with, um, you know, people generating images from ChatGPTs.

There's no consistency between the images for what they're generating. So when we start to break the, the entire workflow down into, you know, really small variables that we have full control over, that's how you can get so much more brand consistency. So yeah, that's, that's just a quick overview of like how we went from beginner to intermediate to advanced with, you know, some image generation techniques that have really helped.

And, um, and the clients are stoked with this. Like this workflow basically generates them hundreds and hundreds of different images for very cheap, like we're only talking. A few cents per image. It basically auto posts into, onto their social channels. And, um, and most of the images are on brand, but even if 90% of the image paid and 10% of them weren't, it's still better than hiring a designer or somebody to work on them manually.

So, um, yeah. Does that make sense so far?

Navid Moazzez: Yeah, yeah, absolutely. This is, this is great and I, I wanna tie this in a little bit to maybe the creators. This was more like e-commerce, this, this brand. But let's say you wanted to use this for, you know, dig digital products or, you know, for your, you know, YouTube or Instagram or whatever.

How would you do it in a similar way to automate a lot of it, or, because it's obviously different when you have a lot of products listing. I mean, that makes sense to, to do it. It's, you're saving a lot of time, but anything like a kind of normal creator, what would they do? Someone who might be watching this?

Mitch Asser: Yeah, absolutely. I, I guess the, the main thing to think about is what do your audience want? So, uh, it, it depends on what you do as a creator. The one that I've been thinking about recently, which I wanna create for ourselves is, I, I definitely believe that AI is going to help solve so many of the biggest challenges we have as humans.

So, uh, I'm, I'm gonna build out a system that has maybe like the top 30 challenges that I believe AI will solve in the next 10, 20 years. And then have an automation that just paints a picture using AI image generation tools of how that gets solved. And it's, uh, it's more of like an inspiration post, like a futuristic post of where we're going.

So it, it's, yeah, it's more of like your visual appeal, very, um, top of funnel wide net approach to try and capture people into your world. So that would be one, um, one way you could, what, what, uh, what other. Uh, types of creators are there, so. Um, I mean, if you're creating digital courses, what would be some good ideas there?

I mean, because AI is getting so good at generating text now, then you could just have some of your, your key concepts, your, um, key objections that you wanna overcome, like headline type of posts getting generated over and over again using, you know, maybe a few different styles. Um, that could be a good way to approach it as well.

Uh, yeah,

Navid Moazzez: I think like mood, mood boards and stuff like that to really make, because that's also a challenge for a lot of people when they create content to have like the brand, you mentioned the brand consistency. Obviously for this e-commerce brand we could tell it's very consistent and it's on brand for them.

I think when a creator sometimes creates content, it needs to be on brand as well. I mean, at least, uh, in, in the style you want and things like that. And I think ChatGPT can be really good to actually have kind of the style and vibe you're going for. For example, now with the creator School brand, we have more of a dark mode kind of obviously suits the AI brand a little bit now, kind of for my, I'm trying because I'm traveling a lot, but I'm trying to kind of make my setup pretty similar.

If you watched. Kind of my first summit. It looks pretty similar. It is not the same place I'm recording from, but it looks quite similar because I'm setting up the lighting in this kind of way, at least. I mean, it's always a challenge sometimes to get it perfect. It doesn't need to be, but at least a little moody and stuff like that.

So I'm trying to kind of get it the way I want. But with AI, obviously you can even, like for example, if I wanted to like take this setup to the next level, I could like take a photo of this. Actually I did it yesterday. I didn't have time, but I could just, uh, feed that into Chachi PT and ask, Hey, do you have any suggestions for this setup?

I didn't do that because then Chachi PT probably is gonna say, Hey, you need to get this, you need to get this. But I can also, I don't have anything else to get. What can I do with what I currently have? And then they will probably make suggestions, maybe to move the light, move the camera, maybe something I didn't think of, uh, to make it even better.

So stuff like that, I use it quite a lot for too.

Mitch Asser: Yeah, absolutely. The, the best part about AI is that the only real limitations that we have, especially when it comes to I image generation, are the limitations of our own mind. Mm-hmm. Like we, we can do so many more things now using AI and coming up with so many creative concepts compared to what was possible just a few years ago.

Like, you know, we've been doing this a long time, Navid, and like, it was so hard to even just create graphics in order to promote the summit back in the day. Now that's like, done so easily with tools like Canva over the last, let's call it five to six, seven years. Um, so yeah, we, if you're a creative person, if you're a creator, then you are a creative person and yeah, you can just come up with any type of wild and, um, creative, uh, strategies and solutions and techniques and just test them out.

Like, you'd be surprised that some of the, the types of images and, and videos that can, that can actually go viral. Yeah,

Navid Moazzez: I think, uh, tools like Canva, they're obviously adding a lot of AI features as well. I mean, they have almost too much, but I think they have to be a little bit careful with kind of where they're going as well, because there's a lot of other players that are coming out that are actually maybe more powerful.

Sometimes you don't even need to edit too much. Like with the, I I just checked out this, uh, image generated Nano Banana from Google. I mean, it's amazing. You can basically, uh, I can upload a photo of myself and tell them, Hey, change the position of this photo. Change me to a different location. I can do whatever.

And it actually has like consistency that is crazy. And I mean, there's tools to to create AI influencers and have this character consistency. I can create the AI photos of myself and turn it into videos. There's really so much you can do. And also I will have some, probably some other videos about this as well, so you can like learn how to do it.

That's something I'm working a lot on these days to kind of see how I can scale some of the photos I've already taken from my trips and like uploading them. And then, yeah, just being creative basically. It's really fun. I don't think I've had this much fun. So it's like just playing around with these tools.

And I know you are more Mitch on the kind of automation side of things too, so you, I don't know if you wanna share a little bit, little bit about kind of the tools tool stack or the AI tools you are using besides ChatGPT? You mentioned Airtable. I think you're using n8n, some tools like that. Anything you would like to share there?

And also you do a lot of things with the Big Moves with the tools. Same as me a little bit. I mean, we have a lot, a lot of AI tools to kind of review and test and things like that. So anything that stands out to you?

Mitch Asser: Yeah. Yeah, absolutely. I've got more image concepts too, if we wanna keep diving.

Well, oh yeah, absolutely.

Navid Moazzez: We have time for that. I mean, I'm sure the audience would love that. It's really good. Yeah, yeah.

Mitch Asser: To, to, to answer those questions. I, I mean, we're, we, we're using basically all of the, the different, uh, chat models. So ChatGPT, Claude, Gemini, um, I do have Grok Premium as well. It's probably one I use the least.

Um, so we're using those for different types of scenarios. Chat, JPT, 'cause it's got a high usage amount. We're using that for most things. Claude for copywriting. It's also better at developments type of work as well. But you quickly run out of credits when using Claude Gemini for like your large context window type of stuff.

So if we have a large, um, piece of code or, uh, like large report that we want to. Breakdown, then we'll use Gemini for those types of things. It definitely doesn't write the best, um, in terms of copywriting and, and the output, but it's good for processing large chunks. Um, we're also using, beginning to use HeyGen a little bit more, doing some testing there with some of the videos though, those videos should be live on our YouTube channel by the time this goes out.

And it'd be interesting to see if people can tell or not. Um, we're not really hiding it. We're probably saying in the Descript, this is AI Mitch, but some of our YouTube videos are me, some of them are, are AI Mitch, um, ElevenLabs. We're starting to use that a whole lot more now, especially with clients. So the, we're using it for some of our YouTube stuff, but also for.

Uh, the conversational ai, uh, chatbots that they have. So we're building out some pretty serious chatbots for clients where, um, you know, they're almost indistinguishable from a sales rep these days with what they can do, can, like, like what they understand the, uh, the, uh, solutions they provide. You can connect them to your calendar.

So they'll automatically book in sales calls or, you know, customer support calls. Um, you can connect them to your, uh, database, your CRM, so you can create new leads, um, and, and have your full call log there. Send through. Those, um, those core logs to either your sales team or your customer support team. So like all of these tools are starting to become really serious, uh, tools for real businesses making real money.

So it's quite exciting. Um, yeah, we've got so many different projects on the go with different clients. Um, and yeah, it's, uh, it's, and then of course we're using the automation tools like make.com, n8n, um, build a lot of things inside of Airtable because it's so flexible. So yeah, they're, they're probably the, the main ones we're using on a day-to-day basis.

Navid Moazzez: Yeah. Lo love it. And with, uh, I think what, what is happening first in terms of like, it's very difficult with the avatar clones to get it like perfect with the gestures and stuff like that, but I think HeyGen is doing quite a good job. It's actually the only problem with HeyGen, I find is the limitation with their avatar four at the time we are recording this one.

I mean, it's, uh, it's really, really good. The new one. You can even have the, the. You can do it now with a video directly to that. So it's really, really good. It's just a limitation of time. So you cannot do too much with their best model, but I think that's gonna get better with time and maybe cheaper. And same with translations and stuff like that.

I think that's also really, really good. Uh, but I think that what, what, as you mentioned with ElevenLabs, I think they are probably the best one. You could, you can basically get it your voice to sound exactly like you if you tweak it. And I mean you can, you need to work with it. You need to definitely do, you need to have a paid plan for a lot of these tools.

By the way, the free plans usually won't cut it to test it. You can use a free plan, but you probably need, if you're serious about this and actually want to level up, you need a paid plan. Unfortunately, the the say is free freemium or whatever, but you can't do too much with the free plan. So with ElevenLabs, you actually need like a creator plan or something to clone your voice professionally.

And then you upload like two hours, three hours of audio. And it needs to be in a specific way. Don't just upload a podcast. I did that mistake myself. It's not good. You need to sit down and read something and like have different, you know, basically reading the way you read and, and do exactly what they say.

And same with HeyGen. Yeah. If I just uploaded this video to HeyGen or like my side of this session would not be good, uh, because I need to do it in a specific way, tone down, maybe some of my movements and stuff like that. Maybe not show my hands as much because then it can be a little bit crap sometimes.

But it depends. I mean you, you can also tell maybe with HeyGen I like that you can, for example, I can hold my DJI mic up and I can do different sets. So it actually looks pretty real. Like I can do a v logging set with HeyGen and it is like, it looks like I'm on the beach or whatever and it's like, I mean, I can be anywhere.

So that's exciting. I think.

Mitch Asser: Yeah, yeah, absolutely. You can even have like a side angle so you can set up your camera, so it's like this, and then you've, you've got your two, two angles as well. So, uh, that's all. You have

Navid Moazzez: 300 looks with HeyGen, 300 looks, yeah. With like, basically depends the plan you have, but you can have like, for sure one, basically like a high quality one.

So you upload that and I'm gonna have some video about this 'cause that's something I'm really, really diving in a lot to. And same with ElevenLabs. I'm, I'm really excited about these creative tools. So it's kind of nice to see the, what Mitch is doing here with the automation, how you can tie into, I think you're using the APIs a lot of, of, a lot of tools to kind of automate and, uh, yeah, it's, it's, I'm gonna check out when you come out with the YouTube videos as well, do you, do you put like B-roll and stuff to, I, I've seen like Annex Ingal, he's doing this a lot and Yulia McCoy, they basically have their clones basically on their YouTube channel and they grow really, really quickly.

But they do have quite some B-roll as well. So is that something you will add in to kind of not show yourself as much on the video?

Mitch Asser: Yeah, exactly right. So at the moment, if we're using any type of HeyGen, then it's a clip that's no longer than maybe five seconds in length. So it, it, it definitely needs to have something else besides just HeyGen.

Like, I, I wouldn't put just purely a HeyGen video out. Um, yeah, I agree with that.

Navid Moazzez: It's probably best. So let's, let's go into a little bit more of the image generation if you have some more examples you wanted to share.

Mitch Asser: Yeah, absolutely. So the, the other big part of, uh, using images inside of chat GPT is actually analyzing images.

And, um, I'm gonna go back to the makeup one in just a moment. Um, but. When by using analyzing images, you can actually do a whole lot of really interesting things at scale. So I'm gonna talk about a couple of client projects here because I find the, these are real world projects, right? A lot of the automation, um, videos that you see on YouTube, they're just kind of made up workflows that people think might work for businesses, but most of the time they don't.

But they look really flashy, so it looks really impressive. But they're not, they don't really have a lot of real world use case show. And

Navid Moazzez: 99% of them show they don't use it. They don't use it. They just like do it for a video, and they have never used it themselves in their business. They're probably, a lot of them have never built a business actually.

They just actually, you do it for a video and that's it. I mean, some of them do have clients, uh, when, uh, there's, there's definitely a few legit ones. But then there are maybe a few kids who never built a business and they just basically showing, showcasing some automations because they're cool, you know.

Mitch Asser: Yeah. But yeah, they go viral. They get a lot of school community members, so sure it works for that. Um, they, but analyzing images, so we have a workflow and that's where it was with a, uh, a, uh, plumbing company. So like they did a whole bunch of plumbing supplies. It's in another country, but our main one in Australia for like our, our big, uh, company here that everyone goes to for DIY stuff is Bunnings Warehouse.

So I'm just gonna use this as an example, but what we were doing with the other company is, so they had like a big umbrella company and then they had like 200 websites with all different niche type of brands, and they were trying to scrape all of the URLs they had from each of those brands and put it into their main company.

Part of that process was scraping images, like these that have. Um, like a plain background. So this one is just the toilet on the white background versus what they called an environment background. So this is the toilet in an actual environment. So we were using, uh, or they were doing it manually, so they have to do like 70,000 products and they're doing it one by one manually inside of ChatGPT.

So obviously that's gonna take a long time to do. So we build a system that does all of that, uh, via an automation, and now they're doing hundreds, if not thousands of those URLs per day. So what it does is what, uh, the AI can do is lemme just download this one and we're just gonna start a new chat and we're gonna upload this image.

And let's see if I've got this prompt in here. Yeah. So this is the exact prompt that we used for, uh, the company that we worked with. And we're basically asking it to analyze this image and tell me whether it has a plain background or environment background. Doesn't make a lot of sense us doing this inside the chat GPT dashboard because we can see that it is, uh, an environment background.

But if you are running it in an automation, then you don't have the ability to look through it and see it. It's just all getting done within the development workflow. So here it's come back and it's told us that it's a IT environment background as we've asked. So we said if it's an environment background, just create MB.

That's how you can use, um, these types of models in order to do some other pretty interesting workflows. So we simply, now, I'm not gonna dive into this too much. This will be more inside of our automation talk, but this is just a quick example of what NAN looks like. Um, these are a couple of the nodes from OpenAI in terms of, um, the, what's available to you when it comes to images.

So this node here is the image, uh, analyze image node. So we would simply add the URL of that image into here. We would then add our prompt into this text input section here. Um, once it's connected, we would choose, uh, the model that we want it to use, and then the output that it will give us over here on the right that we can use in the rest of the workflow would be this output here.

So it'll tell us that this is the type and then this is the, the json URL. Now what we're actually doing in, in the, the reason why we're doing this inside of, um, the client workflow is because we needed to name those images before we uploaded them into the database. So it'd be like MB underscore, uh, 1 0 3 5 7 8, like, which would be like the product number, and then it would number one for that type of image, uh, the first image in that, in that workflow.

So I hope I'm not going, uh, confusing to me with people there, but it's more than anything, I'm just saying that the. The AI can analyze these types at scale as well. So let's just prove this by adding in the, um, the plane background and let's just make sure that it's going to give us the PB instead. So here, this one pb.

So yeah, the, these types of things are now possible with ai, like doing this manually, uh, or doing this previously, uh, using, um, manual processes would've taken so long for 'em to do. Now it's just done within seconds. So, um, that's, that's another way that we can use. Uh, images inside of our workflows, inside of our business.

And then there's a final way, which I'm gonna come back to this makeup brand where it's starting to change now, but what used to happen with, uh, especially e-commerce products or like any type of image that has a lot of detail, is that if it would regenerate the, the actual product, uh, it would get parts of that product incorrect.

So let me just explain that a little bit more before I do an example. So if we look at, um, image here,

sorry, it's just loading.

So if we look at this image here. This product is actually added in, uh, manually. It's not generated by AI because anytime we try to generate it with ai, it will get the logo wrong, it will get the name wrong. The, the extra details down there would just not get generated. So we're like, okay, if we can't regenerate the entire product with ai, what else can we do with it?

And then there's this other, um. Option that they have inside of the API and is also included inside of chat, GPT, called an image edit. So what we decided to do instead is pretend that you've used the product. So, sorry, I'm just coming back into this product. This, this again. So we're saying to the ai, I mean this isn't the exact prompt, but something like, imagine that you've used this product on the avatar.

So the, you know, we mentioned before the, the brand. So it was like a 30 to 50-year-old woman. Imagine she has used that product already and create an image of her with the final, um, product on her fingers, on her face, whatever it is. So that's how we, um, came up with these types of images that allowed us to create AI images that were on brand that made sense for the product, but didn't regenerate the, the product image.

We have a whole bunch of mistakes in it. So we used the image edit feature. In order to achieve that. So you can see all of these, uh, um, I won't click into them, so they take a little while to load, but all of these are getting generated, um, without the, the full product in them, but they still look perfect for that particular product.

Does that make sense?

Navid Moazzez: Yeah, yeah, absolutely. I think, I think this is, uh, the potential or the opportunities are endless. I mean, you can do so much with this, but, uh, what is an, like, this is like more advanced stuff to get a pull all this kind of in, but if some, if a creator wanted to get started with some more, I guess more of the advanced way, like maybe going into some automations, I know you're gonna cover this in the next session we're gonna do as well, but anything you'd like to tease or anything you can share here that maybe they can do if they wanted to automate some part of their business using AI with images or anything else they, they should, uh, they should think of.

Mitch Asser: Absolutely. I mean, the, the most important thing is just beginning to understand the capabilities. So that's why I did, you know, I've packed a lot of different concepts and, and scenarios into, into this, um, talk here today to like, just show you what's possible. And ultimately it's just under, once you understand what's possible, then you can start to think about your own creative process or your own business and figure out what parts of, um, the business you want to then automate.

So, uh, yeah, the, the next session that we go into, which will be more around automation, will be, you know, a, a scenario that most creators can relate to, which is if you can create, you know, one long form piece of content in your own words. That, uh, maybe it's a YouTube video and you pull the transcripts, maybe it's a long blo, uh, newsletter, and you pull, pull that into a, um, an automation.

I'm just gonna show you how you can take that and then turn that into dozens of different posts, maybe hundreds of different posts, basically on autopilot. And there's a few really interesting techniques and, um, different tools that I found in order to do that using existing viral templates and those types of things.

So, um, yeah, that, that next automation talk is gonna be, um, it's gonna be more into the, the weeds of like NAN and how do you build these workflows? How do you, uh, make, how do you pull information from Airtable? How do you send it back into Airtable? How do you make it go through the APIs? So like, yeah, that's gonna be a very interactive, um, talk as well and super valuable for sure.

Navid Moazzez: Yeah. I lo I love this because like as you and I, we have me, you do your YouTube videos. I do now, for example, this summit, uh, or, or any, any sessions, basically any videos. A lot of creators are creating this. I mean, they, I think they are missing to repurpose it in on a lot of platforms. And I've done this mistake myself before, like I had so many sessions, but because it's so time consuming, a lot of it, you never get around to it, basically.

But with AI and these automation tools like NA 10 Airtable, then you can do so much with it. And I think that would be kind of interesting to see how we can maybe take something like we have done here for, with these videos and then repurposing that into a lot of different pieces of content. Because from like, uh, 45 minutes to an hour video, that's a, that's quite a lot of content.

That's a lot of words spoken. So, I mean, how many pieces of content can you realistically create from something like this, Mitch?

Mitch Asser: Uh, I mean, it, it's, it's ultimately, it's limitless. Uh, and that that's the, the, the point it the way, the way that I approach it. So, um, just a another, another quick teaser is we've got like 200 different viral posts templates laid out, and we take the transcript and run them through all of those.

So I'll create all 200 in one go because it's so cheap now in order to do so, and then you just start culling all the ones that don't make sense. So you might end up with 20, 30 really solid pieces of content from that 200, but it's still so worth it. It might only cost you a few dollars in order to, to create the, the, the text component for, for those posts.

So, yeah, I, I, I think that's, um, that's definitely possible. But it, it really comes down to a whole bunch of things. So you mentioned there like so many creators don't recreate on all platforms. I think there's a time and place to do it. You, you don't wanna be. Trying to do that from day one, like at least get into the flow with one platform first and then start to look.

But the second really big benefit from using automation tools, um, from building out your own workflows is that now you can customize the repurposing of your content in order to fit the current workflows you have, the current tools you have, you're not actually introducing new software where you have another dashboard that you have to go and use.

You can actually, you know, if you are using Notion, you can build a workflow that's generated via notion. If you're using Airtable, you can do it. That's generated via Airtable. So there's yeah, that the ability to repurpose Content at Scale using the same tools that you're already in on a day-to-day basis, that I think that's the biggest selling point of automation.

Navid Moazzez: Yeah, because it's interesting. I'm using obviously Notion a lot. I'm more of a notion user than Airtable, but I, I do like both. They're good for maybe slightly different things and things like that, but the notion now in, I mean I, I think Airtable as well, but with CLOs, cps, and the connectors you can have there.

I mean, it's pretty amazing. Like if I wanted to have a speaker list about ai, for example, it's not perfect, but I can basically connect notion inside of Claude, go in there, even ChatGPT. But I have found issues with ChatGPT. They don't wanna, I think they cannot like create the databases as much in, in Claude.

They can actually create it for you like a mini agent. It's not kind of like the easiest way to set up an agent is actually to go to your favorite, you know, AI LLM tool like Claude is what I prefer for this, it's the best one probably for this use case chat is. Okay. But, uh, with Claude I can just, uh, connect it to Notion and then it creates a database for me with pretty accurate results.

It's pretty decent. It's not like maybe all the speakers I would consider, but it's quite some, I even saw some speakers I reached out to and some maybe I wanted to reach out to anyway, and some I didn't consider. So it's quite good. And it's like basically doing that in a few, maybe a few minutes because it needs to think a lot and process, but it's like categorize it and everything and that's.

Probably one of the easiest use cases to just get started with a little bit automation. You just have to click one connector. You don't have to like deal with, before you had to actually deal with like APIs and CPS and all this. Now you just connect it and it's there and you just, uh, toggle it on. And you can do the same for Stripe, PayPal, uh, there's quite, there's quite some for I think your Google Drive to access all your documents, for example.

And I could, and you mentioned like with creating dashboards and stuff like that, I think you are doing some, some cool things in Claude. You can do that very easily if they, if you wanted to like analyze your sales and your business, you can do that in Claude, like really easily. You can just Claude, look at my Stripe dashboard.

Obviously you have to feel comfortable to connect it, but you know, they cannot steal your money or something like that. I mean, it's pretty secure. So I would, I would just encourage you to try it and you can create a very nice dashboard. It's like in a few minutes, it's, it's all there.

Mitch Asser: Yeah. Absolutely. MCPs are great.

We, we even have some where you can paste in a URL from Twitter for a video and it'll download the video onto your, um, onto your desktop and yeah, there's so many, so many things like MCP is a whole more whole like bigger talk and advanced, uh, session. And no, with, with the

Navid Moazzez: and and attendance stuff, you can do so much 'cause they have so many different things to connect.

And I know you have used, what do you, what do you use most? Like, make is Make is pretty expensive compared to and, and eight 10 I self hosts it personally with hosting Air. I don't know what you guys use, but I use a hosting air plan with their, they have a special one for N eight 10 and I just hosted like that.

It's like really, really cheap. And, uh, obviously they have some benefits on the ho on the N eight 10 website, but I, I would recommend just the hosting or one because it's so cheap and, and you can just have basically unlimited automations. You would probably not run out anytime soon with, uh, to upgrade your hosting.

You can just upgrade your hosting or plan at some point, but I haven't needed it so far.

Mitch Asser: Yeah, we, we have both. So we have the hosting a plan, um, we don't really use that one much anymore. We actually use the Claude Instore because the Claude now has community nodes, so you get a whole bunch of, um, extra nodes from, built from some of these software tools.

So instead of having to figure out the API docs yourself, then there's nodes ready to go. And so I, I always updates. Updates

Navid Moazzez: is annoying, man. I hate the updates. You have to like update it and it's like you have to go into the terminal. So if you are not so techie, like you don't need to be techie. You can figure it out with ai.

I'm not techie with this stuff or like. It still took me some time to update the stupid thing with NA 10. So like with hosting or, it's not that easy to do it, so you have to kind of look into it. Yeah, and the, the support is terrible. They cannot really help you, so you have to actually figure it out a little bit.

But it is a lot cheaper than the Claude, Claude version. I think, uh, at least if you're, if you're gonna try it out. So, I mean, for me it's okay.

Mitch Asser: I. The way that I think about it is this, you're either paying in time or money, right? So like you could get the cheaper version through hosting up, but then if you are building a workflow, you might have to read documents, set up a HTP call.

So you might spend an extra 10, 15, 20 hours per month building nodes that are already built in the community nodes, um, section. So if your time is worth more than. $2 an hour, then it probably makes sense to actually have a Claude account where it's just gonna make things so much easier for you to, to build.

Um, yeah, you're always paying in time or money, so be careful which one you choose. What plan

Navid Moazzez: have you, have you guys been, because you use a, use it a lot, do you have like a very high plan because you do it for clients or, or No. Needed,

Mitch Asser: I mean, uh, most, most clients wanna work in, want us to work in their accounts, so it's easy handover and we we're not running a huge amount.

Like, um, the, the, the, the intro, uh, account for NAN is pretty, pretty decent and they're actually improving it over time, especially. So, um, Google actually working on a very similar type of tool now as NAN. So as more competition comes up for NAN then the, the pricing will go down for sure.

Navid Moazzez: Yeah. The reason I think I picked the, basically the self-hosted version is because we are using it for.

Email replies and you know, when you get a lot of replies like that, it's kind of, it's kind of adds up. And that's why I couldn't use make, I, I just blew out the plan immediately. Basically for my line. I tried make for this was, it was too expensive, it was not making sense. When you get like thousand replies or like more, very quickly, like 20% because everyone basically, it's like to reply to get the lead magnet is one of the best ways actually, it's really easy to set it up with like, because you wanna have it as a personal email.

So when they sign up to your newsletter, you have a trigger word basically for, uh, yeah, I mean basically when you sign up to this, probably you saw it. So you have like reply AI tools or something like that, and then you get this immediately. But it's a personal email for me basically. And that's set up.

So you, you can't really do that easily with Kit ConvertKit. You can't, it's pretty difficult. So you have to set it up with uh, Gmail trigger and then you can tag them still inside of your email provider like Kit. In my case. And then it's all organized and things like that, and they're getting what they ask for very easily.

But, uh, when you get a lot of replies, it might get up there. So that's why probably I pick the self host, maybe the N eight 10

Claude

Navid Moazzez: version is okay. 'cause I think they have a much, much higher limit than, for example, make.com, which is, uh, yeah, they, they have more, more limits early on there for the pricing plans.

So Anything else Mitch, you wanted to share here before we kind of wrap this one up?

Mitch Asser: No, I think this has been a, a great session for sure. But if there are any questions, and I know we're gonna be going live each day and, and diving into those questions as well, so make sure you attend some of them. We can answer questions about this or about any of the other talks that are out there, or just AI questions in general as well.

I'm excited for, for those live sessions.

Navid Moazzez: Yeah, absolutely. And where can people connect more with you, Mitch, where is the best place? I know you have Big Moves. Any, anywhere else you would like people to connect with you?

Mitch Asser: So if you wanna dive into AI tools, reviews, uh, go to BigMoves.ai. Or if you are somebody who's a more experienced business owner looking to automate workflows and need someone to help with that, then you can go to MakeBigMoves.ai and see all the details there.

Navid Moazzez: If you found value from this session, let us know in the comments and you know, we will definitely try to help you as much as possible. It's very exciting with all these tools and how you can use it. And the most important is that you dive in, you know, the AI is not going anywhere. You need to use it in your business to basically amplify yourself and what you're doing so you can scale and be relevant.

Because a lot of creators and entrepreneurs who's not using AI, they will be obsolete pretty quickly. Maybe in the next year, 18 months or so, they will be wiped out. A lot of industries will be wiped out eventually. Like even with filmmaking and stuff like that, it's like obviously they're very creative so they're kind of safe for now a little bit.

But if you are using AI, you can charge a lot. I saw a guy, as I mentioned earlier, it was like 50,000 to a hundred thousand dollars for like an AI ad using Veo 3. He is good with storytelling and video, but like you can use a lot of these tools to charge really good money if you wanna work with clients, or you can use these tools to basically scale your own personal brand and business.

So lots of opportunity and I'm excited for your next session, Mitch, as well with automation. That's probably what I'm really, really excited about diving into and some more technical details. So stay tuned for that and I'll talk to you guys very soon. Ciao.