FREE n8n AI Ads Generator: Nano Banana + GPT Image 1 = $21k In 1 Week (AI Agency Meta Ads)

Last week, I made $21,600 for my AI agency using AI generated ads. No cold calling, no cold emails, no Upwork or any other BS method. This came from cold prospects who had no clue who I was or my company. And this didn't come through my YouTube channel or my personal brand. And yes, of course, we're scaling this method aggressively. Now, as a matter of fact, I or no one from my team even spoke to any of these leads to convert them

Into appointments. I also had a voice AI agent do so for me. Now, one of the reasons I was actually able to do this is because we are now able to generate unlimited creatives in seconds. And in this video, I will show you exactly how you can generate thousands of creatives in 5 to 10 minutes with a few simple copy and paste prompts. Now, I'm just going to walk through a quick demo right here, and you'll see some other examples as

Well in this video later on. Now, I also want to show you some variations here for the generations. So, I changed up the headline here just to join our stride AAI Academy for free. And you'll see that pretty much everything else is the same except I switched out some of the styles. So, some of these styles are all unique. And then here, I just wanted to show you the memes particularly. So, um, and you know, you could use memes, you can

Use different things like you guys can get creative here, but you can see like we got the hotline bling meme. Uh, we got guru's plan. These are just different creative things that, you know, the AI will actually take into account for generating these images, right? So, mocking Spongebob, uh, Batman slapping Robin, uh, the two buttons meme, and then Oprah, you get a car, right? So, if we go ahead and generate all these, we'll see what we actually get. And then for

These last two, I'll change the variations to four and five, just so you can see that the variation system we built actually works. We can see all these executions now are running. So, looks like everything is going good so far. All right. And now we are seeing some of the generations come through here. So, we can see Gru's Plan and Mocking Spongebob. This one's in manga. Uh, this one's in Claymation style. Here we got Studio Gibby right here for Oprah Winfrey.

You want a car. We got Dragon Ball Z here for the Batman slap. So, if I click on some of these, um, we'll see their different variations. So, join the Strat AI Academy for free. Join the Strat Strat AI Academy for free. We click over here. We got these ones of Spongebob with manga. If I go here to Batman, we can see this one and the two buttons right here. We got this one in Rick and Morty style, right? So, you

Can see the different styles here. And it should generate four like you can see there. We go to this one, Oprah. We can see that she's doing the winner car sort of style right here. Um, so you guys get the point. Uh, this is pretty damn cool. This allows you to create so many different variations, so many different types of creatives for really any type of ad or really any image you would like. Now, the stack that we will be using

In today's video includes Nano Banana, Google's new image generation model, GPT Image 1 by Open AAI, and we'll be using this on Kai AI, which I'll show you in just a second. We're also, of course, going to be using N8N for our workflow automation. We're going to be using GPT and Claude for the LLMs. And for logging our ad creation, we're going to be using N8N's new data table feature. And if you're not familiar with that, I'll dive into that briefly

In just a second. And this is what the complete workflow will look like. I want you guys to know I spent quite a few days building and refining this entire system that I actually use in my business. This isn't just some template or some copy and paste nadn template that you may see on other YouTube videos. I actually currently use this in my AI agency to make money and scale my business. And before I dive in, you can get this entire

NAND workflow for free by joining my new school community that I'm launching right now, the Stride AI Academy. Link for this will be in the description, guys. I am going to be putting so much value into this community. We're going to have many different videos on NADN automations, Stride AI tools, different CRM setups, uh, AI coding, a bunch of different things, but specifically many different NHN videos because this is one of the main tools that I use in my AI agency

Today. Now, before I dive into the workflow, once again, this is really how this system works. AI powered meta advertising system. So, the AI ads that we're going to generate here. And then we utilize instant lead response AI systems. So AI appointment setters via SMS, email, chat, WhatsApp. More on that in future videos. We're really going to be focusing on the AI ads here. And these are of course static images just to show you the proof right here. 21.6K Canadian right

Here in my Stripe account. And this is for a week time frame right here. And you can see the cost per leads that we were generating with some of these different ads, right? So this can vary. It can go up or down, but this is just a sense of what we are getting. So, here's a preview of what we're going to be setting up in today's video. So, not only do I give you access to the N8N workflow, I'm also going

To be giving you access to this Air Table base. Now, what does this base entail? Well, this is really where we actually uh generate these ads and these creatives. So, you can see here, these are just some different ad creative examples that we've been generating. I've been generating uh really hundreds of different examples probably like about five to 600 maybe even a thousand here because a lot of these are you know variations of eight or four or whatever the case may

Be and I'll click through some of them to give you a sense and we're of course going to do some test um executions in just a moment but the reason this works so well is because there is many different dynamic features which you'll come to see in just a second and you can really customize this for really any which way that you want. Even if you're not going to use this for running ads, let's say, maybe you just want AI image

Generation. Well, you can use this base as a framework and plug and play your own different ideas. Um, which I'll show you in just a second. All right. So, here I'm actually going to show you an example of me generating an ad. So, first things first is we have our headline, our big claim/hook. So, this is just very basic. I personally wouldn't use this, but we have streamline your business workflows with AI automation. So, I definitely, like I said, if you're

Actually running ads with this, I would make something much more of a promise, something more of a hook that actually gets people interested, you know, outcome driven. Next, we have the sub headlines. So, supporting proof/context. Very, very basic. I'm just keeping it generic for this example, guys. So, definitely get creative here, guys. I will be making more videos in the future on actually how to structure your AI offers. I'll be posting more content like that as well in the school community.

So, make sure to join. But we're just saying right now the future is here. And then for our CTA, we'll just say learn more. Okay, so this is the CTA call to action. You can really customize any of this. Now, you don't have to fill all this out. You could actually fill none of it out if you want. You could just fill out the prompt. Like all this that I'm showing you, you don't need to fill out every single thing. And

I'll show you just in a second. It's dynamic. Um the way we set it up. Next is a prompt. So if we want to use this, we can. So I could say I'm just going to say like make it kind of comedic. You can really say anything in the prompt. You could say, you know, do a comparison, do really whatever. Right. Next is what I call foundations. It could be called offers, whatever the case is. It's linked to this table over

Here. And this table is really just, you know, whatever offer you're running. So, I have one offer right here. Um, then I have some offer details. This is like a short summary of the offer right here. And then right here, I have a foundational dock. This is more a little bit more proprietary. We just have a long dock right here. Um, and then if I click generate, it actually just makes a summarized shortened version of the offer. But the main thing

You would want to fill out is just the offer details. So, information about your offer, maybe who you're targeting, really anything. It doesn't even have to be too in-depth. And you, like I said, you don't even have to use this. It is an option, though. Next, we have some styles. So, you can select from different styles here. So, you'll see here that I have quite a few. I have like Pixar, um, Disney, Renaissance, uh, Simpsons, you know, South Park, you know,

Dragon Ball Z, Future Rama, like a bunch of different ones. And you can really just go to your styles tab here and just fill out a bunch of styles, put some style prompts. You can get this generated by Chad GBT. Put that in there and you can use some of those because what we're doing here is in our images table. Let's say I select a style like Pixar. I select my um foundations offer right here. And you can see that we

Have a lookup record for offer details as well as style prompt that's being pulled in right here. And this will be referenced within our NAD workflow section. Next, we have a memes tab right here. So, I could select a specific meme if I want, like Spider-Man pointing, Drake hotline bling, distracted boyfriend. You don't need to do this, like I said, but if you want to like insert some memes into your uh AI ads, you can definitely do so. All right. Next

Is where we're going to select our model. So, we can use either use GPT or nano banana. I'm going to use GPT for this example here. We can select our uh sizing right here. So auto 1 3 4 9 to 16 4 to 6 16 to9 you really have different options. I'm just going to keep it 1 one. If you don't put anything here it will default to one one and then variations. So you could do like one you could even

Add like even up to 100. You would just spend a lot on tokens here but I have it up to 10. I wouldn't really suggest any more than that. I would just if you're going to do more just do a different generation. U maybe add some tweaks to it. Trust me on that. Um, so we could do like in this example, let's do three. All right. And then all I have to do is click this check mark right here and it's

Going to actually launch the execution. So we should see this send any moment now. Okay, perfect. So we see that the execution is running right here. Okay, so the execution succeeded. So basically we just went down this path right here. So, I'll break this down more in depth in just a second, but you can see we have our GPT um section right here. And then we have our nano banana section. So, depending on what it is uh for the model that

We're using, it will go down each section. Each section is really the same. You'll see over here, I kind of noted what these what these different paths are. So, first path up here is if there's two reference images included, and I'll show you that as well. Uh second path is if there's just one reference um image included and the third path which we used is if there's no reference images included. So you can see that since we use GPT it went

Down the third path for GPT which is no reference images. Okay. Then it uh went through the LLM. It used NADN's data tables right here to log this AI generation. We had a bit of code here and then we're looping um which is contingent on how many variations we want to create. Since we wanted to create three, it looped three different times. Use the Kai API to create the image. Use the Air Table get record to get the record, update the

Record, and then wait a little bit and then just loop right through. Now, when each one of these loops for the AI generations are done, we're going to get some executions right here, which depending on if it's Nano Banana or GPT, it's going to fall into one of these. These are our callback URLs right here. And we're just searching for that Air Table record and then adding the image to Air Table via the API, downloading the image as well, and then

Uploading it to Google Drive. All right. So now we can see that three executions just popped up right here, which means that they are done. So we can see this successfully executed. Now, if we take a look at our Air Table base, we have three images here that should be kind of similar to the distracted boyfriend meme with a Pixar style. Um, kind of comedic and they are talking about AI and AI workflows. So, if we take a look here, we

Got streamline your business workflows with AI automation. The future is here. We can see the girl, the guy, he's looking at the AI robot. We take a look at the next one here. Streamline your business workflows with AI automation. future is here once again. Um, following all the different prompts that we gave it. This one looked like I messed up here, but you get the point. You can really generate many different types of variations for your images. I'm not going to

Go through them all, but you can see here we got some Dragon Ball Z ones, some South Park ones, different ones like Kermit the Frog drinking, different memes like that. A bunch of different varieties. Now, you'll also see here there is a option to add images. So you can add reference images into you know your uh generations. So you could either add one or two. So with this you could do things like for example combining two images together. I could take

A picture of me, take a picture of you know um Hawaii and say put me in Hawaii for example. So you could do that or you could just add one image like you see here. I added my logo but you can add different things into your images as well. So, you don't need to use these, but like I showed you in the actual workflow right here. If you use two images, it's going to go through the first path. If you use

One image, it's going to go through the second path, and then no images goes through the last path. So, the second path is actually this one right here. The first path downloads both images here, and then goes on. All right. So, to show you one more example here, I'm going to generate a bunch of different ones right here. They're going to use the same headline, CTA, sub headline, etc., as well as the prompt, but we're going to use different styles. So,

Here I got Simpsons, Family Guy, Dragon Ball Z, um, Studio Gibby here, a bunch of different ones, Futurama, and we're going to use some that are GPT as well as some that are nano banana. We're going to use one by one for most, and I'll use a 9x6 as well. And then we're going to do two variations for each. And I'm going to generate all these right now. Now, you'll also see for some of these I'm using a reference image. So,

Here I'm using one reference image for these two. This one is Executive Stride logo. This one is just a picture of me. And then here for this one, I'm using a picture of me and the Executive Stride logo. So, I don't know how these will turn out. I haven't really done too much experimenting with actual human photos here, but we'll see how it uh goes right now. So, we can see all these executions running right now. Okay. And we can see

That the ads are already starting to be generated. So, if I click into some of these, the Nano Banano got done really quickly. So, these ones go by pretty quick. And we can see streamline your business workflows with AI automation. Future is here. Studio Gibby right here. And we got another one like so. And if we go to this one, which is supposed to be kind of Futurama type style, you can see we got Bender here. And we got more right

Here. So, um, GPT is still generating. Okay. And now you can see these GPT ones are being generated. We got the Simpsons right here. So, if I click into this, we can see streamline your business right here. We got the little Simpsons AI bots right here. Um, and we even got some we got me me right here. So, this is my picture. And we can see that it is this is me when I was bald. Um, and this is me again.

So, not perfect, but you can do some tweaks here. We got even more right here with uh myself as well. Okay. In the house. And this actually looks I mean not exactly like me but it could take a lot of different things from the image and it even has our executive stride guy right there. So pretty cool stuff. So I pretty much just showed you every single example that you know these different flows could go through. Now I'm not going to

Walk through each and every single one. I'll mainly walk through you know some of the main ones. Uh these are really duplicated so keep that in mind. But let's start from scratch. Um because you can either go ahead try to build this out um but like I said join our free school community. You'll be able to get access to this workflow 100% for free. This is something that I've used to personally make money thousands of dollars like I showed you. So

Yeah. So what we're actually starting off with is just a basic web hook trigger. So um you're going to make a web hook trigger. And what we're going to do is within our air table base, you'll see under automations, what is happening is when this record is checked. So when we check that box, it's being triggered. And then we're running a script, which all you're going to have to do is just come into this script and simply just change out this

URL for your production URL of the web hook trigger in N8N. Okay, so that is going to send to N8N and begin the actual workflow. Once that happens, we're getting the record that we want simply by just grabbing the record ID right here. And then we have a switch node right here, which is basically just checking to see if we're using a nano banana model or if we're using a GPT model or if the uh model doesn't exist. Because if the

Model doesn't exist, you can either put it to nanobanana or GPT. It's kind of just your default one, whatever one you want. Right now, I guess I have it set to nano banana. Honestly, I would probably switch that to GPT just because personally I've tend to get better responses with GPT, but you can use nano banana or GPT, either one of course. Right. Next, we have another switch node. So, this switch node, first one was just for the model. This one

Is checking to see if we have an image. Right? So it's just looking to see if this image field exists, right? That means we are using images or if it does not exist. So if it exists, that means that we're either using um you know, one image or two images, right? And if it doesn't exist, then we're just going down here. So this switch node right here is checking to see if we have the image two that exists or not. Right?

If image two exists, then we know that it's a two image generation. All right, we don't even have to check for image one because we know that you're going to be using image one and image two. So then it goes down this path right here, which I'll go through in just a second. If image two doesn't exist, then it's going to go down to this switch node, and this switch node is going to check if image one exists. If image one

Exists, then that then we know that we just want a one image reference generation, which is this path right here. And then, of course, if no images exist, then we're just going to go down through this path right here. So, I'll walk you through what image two uh with two images looks like. So, what we're doing is if it, you know, image two does exist and we're going to download image one, right, with just a simple HTTP get. And then same

Thing with image 2. We're just going to download that. And once we download that, we're actually using imageb.com. So you can just go here, sign up for free. This is actually just a place where you can upload a binary right here to be able to host your image um online for free. All right. So we're just calling that API. So simply just go there, sign up for it. It's very basic to just get it set up. And then what we're doing

Is we're using a merge node right here to combine these two. And then what we're doing is we're using a GPT node right here to analyze image one. So all I'm saying for the text input is what is this image? Okay, it's going to give us, you know, really what this image is. And then we're using another GPT um node to analyze the second image. So what is this image? This is just going to describe what the image is for us.

Okay. Then what we're doing is we have a AI agent node right here. So you'll be able to see that this is our prompt. Now I'm not going to read through this whole thing. You can read through it if you'd like, but it's basically just saying, you know, you're an AI image generation prompt expert. This is what you specialize in. It's giving some context like always make sure the text fits on the image and does not get cut off. This is

Very important. It's telling it what its task is. Um some basic um you know points around prompt handling. If a prompt exists, then expand it uh while preserving its core intent. If no prompt exists, create an intelligent prompt that combines the analyzed image in the most effective way for advertising. And then image analysis integration. So use the image analysis data to understand visual elements, composition, style, and content. Incorporate relevant details about lighting, mood, objects, people, text, and overall aesthetics. Okay, you

Can just go through read this. Now, also do color integration. So, if you do want specific color schemes, you can actually add the hex codes right here. So, there's options for three different color schemes. So, I didn't add any of but I've tested this. It works. You can like add hex codes here if you'd like. So, if it does mention one, two, or three colors, then you can use the those colors as primary, secondary, and accent colors. But if it doesn't

Mention them, then you don't have to use and you can just, you know, use whatever colors. Business inter intelligence integration. So this is referencing that foundational offer information and how to actually use this pricing guidelines. So here I just state like never mention any dollar amounts or pricings or anything like that. Output requirements. So some stuff here quality standards and then response format. And then we're referencing our business intelligence info right here. So the offer details, the prompt, the image analysis

That we got from here, the image analysis two and one, and then the different colors if we use any, and then the style. So this uh image generation should be within this style. And then we're mentioning, you know, the headline, secondary headline, and CTA, what they are and how to actually use them. And then we're mentioning if we reference any memes. Okay, so I know that's a lot to kind of take in. You really don't need to fully like re understand

The whole thing. You could just use this out of box. But if you do want to go in here, tweak some of the prompts. You can definitely do so and optimize it for your use case. The one thing I will say is depending on each path, the prompt is a little bit different. This prompt right here has everything because it has um the two image generation analysis information here. Uh the second prompt that only has one image analysis, I think it's

A little bit has a little bit less. And this one here even has less because you know it doesn't need any images. So yeah, just if you want to like look at all three of them, you can kind of see a comparison, but they're all other than that it's pretty much the same after that. Um and as a side note, we're using for the chat model, you can use anything. We're using GPT5 mini right here. So we get good quality at

A cheap price. And then as a backup, we just have Quad Sonic 3.5. You could really use anything. I have a calculator here for no particular reason. Then we have a code node right here where we're basically um just returning some dummy data. So depending on how many variations we actually set in our air table base, like if we set 1 2 3 4 whatever um if we set 5 10 you know 15, it's going to generate 15 dummy datas or

Five dummy datas. Right? The reason we do this is because if we go into Kai here, like I mentioned at the start, this is the model that we're uh the the uh tool that we're using for these AI image generations. So, it's actually a very cheap platform. You can also, you know, switch this out and use other platforms, too. There's many different ones. You could use foul.ai. Of course, you would have to change this workflow, but for this workflow, we're using

Kai um AI. Um there's many different ones. You could use the direct API of Google or OpenAI. But yeah, so access the best video, image, and music models in one API. So it's pretty um convenient. You can see we got like many different ones. V3, runway, seed dance, and you could actually improve upon this workflow. And if you guys do, make sure to let me know in the school community because I'd love to see it. Um but we also have all

The different image generation models. Nano, Banano, Seeddream, um 40 image, uh Flux, IMI, uh idoggram, Midjourney, Imageen, Quen, a bunch of different ones. So, you could add uh additional ones as well if you want. So, sign up for here. Um buy some credits. It's pretty cheap, very cheap, actually. And then just grab your API key, and I'll show you where that node is in just a second. Um but yeah, back to this. So depending on how many variations it's going to

Generate that dummy data because like I was mentioning in Kai over here um a lot of these like I think 40 image has a it does give you the option um for to send specific variations but it it's only like one two or four. You can't do three or any other number above that. So that's why I just thought it made more sense to just have a loop right here. So we can really give it any amount of variations. And then

Uh nano banano I believe in Kai doesn't even allow you to put any variations through their requests. So that's why we're using the loop. So we're sending multiple variations and we're just doing so by creating the dummy data here and then looping through however many variations we put in that air table base. Okay. So I hope that makes sense. After that, we're using NAND's new data tables feature. This is a really cool feature that NAN just came out with, and I

Won't have time to go through a deep dive of data tables in this video. I'm definitely going to do more videos on data tables. And this is a beta feature that actually was released last Monday and now it actually just got fully released onto the um self-hosted version, right? So, it got released here literally just 7 hours ago. So, so if you're not familiar what data tables is, it's essentially like spreadsheets but a lot quicker. You don't need to fumble around

To connect your Google account through authentication or anything like that. It's directly into NAD. Now, here you can see this is really just a basic data table. We're really just logging every single time we create an ad and we're just really um putting that same information that we are, you know, in entered into Air Table into our data table. Um, you know, we don't have every single field. We really just have the main ones that we inputed. So, you know, our

Headline, our sub headline, our uh prompt, CTA, foundational information, the model we use, the sty uh size, offer details, style prompt, record ID for our air table base, how many variations, the style, and then the created at and updated at. You can add different columns here, strings, numbers, booleans, date, time, and this is still in beta, so they definitely are coming out with a lot more upgrades to this to make it a little bit more production ready. So, for the time

Being in this workflow, we're really just using it for logging capabilities. Um, but if you didn't want to use this for whatever reason, you could just remove this and the workflow would work the exact same. All right, so just so you guys can see here for the image analysis, I just resent an execution here. So you can see the image shows a man standing in a modern well-lit kitchen. This is the picture of me with a beard. Gives the uh analysis

Right there. Same thing for the logo right here, executivestride.com. Just so you can get a sense of what that would look like. Next after the data table is we have our loop. So it's just looping right here with a batch size of one. All right. So this right here is where we're calling the Kai API. So, as you can see, this is going to generate the image. Um, what we're doing right here for the nanobano is we're calling this to create

A task. We have our Kai AI um in the bearer off right here. And then here in the JSON, what we're doing is for the model, it is nano banana. And then for our callback URL, this is where you would actually want to put that URL that I'm going to show you in just a second. The one down below. You're going to put that here for your call back URL. The prompt is just right here. You can see this is the

Prompt that GPT generated us and then the image URLs are right here. So this is where we're actually referencing the image URLs in um 2Kai because not only are we just an analyzing the image and giving that to GPT for the prompt because we want the prompt to kind of take in what these images are about. We're actually able to just upload the images too of course to uh Kai. So we're doing one and two since this is the two image

Path. And then our output format that we're doing PNG and then the size right here is one one. So of course there are some different um parameters you may be able to pass. You can change this to JPEG if you want. Just check the Kai docs if you want to do anything different here. But this is really all you need to do. If everything goes well you should get a 200 code for success. And you'll see here that we're getting a

Task ID and a record ID. Okay, the task ID is important. This is actually what we're going to be using. So with that, we're actually getting that same record ID that we got from the very start. All right, we want to get that initial record. And what we're going to do is then in this node, we're updating that record from our same base with the task ID. So we're adding the task ID here. Um, but you'll notice since we got the

Record already in this node, we're we got the record and then we're adding the old task ID that we may have had. If it's the first generation, it won't be anything. But let's say it's the third generation, then you're going to have two task IDs from the previous one. So, we're referencing those and then we're adding in the new task ID that we just generated from this request. All right? So, this is why we're looping. uh each loop we create the

Image and then we're adding the old task ids. So let's say we generate 10 images, we'll have like nine task ids at that point and then the new task ID from that other uh recent image generation. Okay, then I just have a weight right here of 5 seconds and then it just loops through. Okay, the reason why we add all the task IDs is because since we kind of like you know made our own variation um system, right? uh to be

Able to do like one 10 gener variations depending on how many we want. Um when we come to our uh fallback URL here and we're we're searching a record, we're going to search the record based on any record that contains the specific task ID um when the generation hits. Okay. So what I mean by that is each time it loops through this um trigger down here is going to be hit every single image, right? because that's the fallback URL that we

Defined in our call request. So if I grab this data right here, you'll see so here we're just getting that information from the web hook. So this is the image when it's done generating it's going to trigger the callback URL. You'll see the completion time, the cost time, uh created time, the model, the parameters, so the input output and the result in JSON format. And then what this right here, this code snippet is doing is it's just pulling that URL out

From the uh JSON results. So we got that URL right here. So if I went ahead and searched this URL, you can see we got our image here. This is using the nano banana one. So like I said, sometimes it's not as good that I find with Cernet images. All right. Um but you can see that definitely does look like me. And then what we're doing is we're searching the record right here. So we're using a filter by formula and we're

Just finding that record that contains the task ID. Okay? So where the task ID contains this and then we got that record because no matter if it has 10 task IDs in it, if it contains this new one that came in, it's still going to be able to find it. So that's what we do opposed from just like updating the task ID and replacing it with a new one because then it won't be able to find it won't be able to

Just add all the images into that same record. it would just replace that image. Okay, what we're doing now in this code block once we get that record is we're grabbing all the URLs for that record um of the images. So if it has one image, if it has five images, we're grabbing all those records as you can see right here. And if I go into JSON, like it's literally all images and then URL for the different ones right here. And

Then we have an all images string right here. And this is what we're actually going to be using. So if I go over to the update image right here. So what we're doing in this uh request, the reason why we have to use the actual API and we can't just use the air tableable node is because it doesn't have what we the specific action that we want here. Um we actually just want to update an image and add it onto it.

Um so we need to use a patch method right here. We're calling the air table API and you'll see that we got you're going to need to change these two out the app ID and the table ID. Way you get this is you go to air table and you'll see here that we have an app right here. The app ID is this right and the table ID is this. Now make sure when you need to be in the images table when

You get your table ID because if I switch a different table you're going to get a different table ID up here. The app ID is just the whole base. Um but the table ID changes. Okay. So, make sure you get the table ID here and the app ID here. And then you add those in right here because that's what we're going to be using to then grab the record ID from that base. And then what we're doing is we're passing for

The JSON the AI images right here. And this is literally um those all images strings right here that we're passing through. Okay. After that, we're literally just downloading that new image and we're simply just uploading it to our Google Drive. The reason I upload it to Google Drive is because Air Table is a nice way to visualize the new ads, but in terms of downloading those new ads to be able to actually use them, it's a lot easier to bulk download

When it's in Google Drive. So, that's why we do it in Google Drive. Now, I'm not going to walk through the fallback URL process for the GPT. There is a little bit uh it is a little bit different just because the data that we get from Kai is actually a little bit different for the nano banana versus the GPT. So, some of these code nodes are a bit different, but it's basic. It's the exact same process. Okay, we're doing the exact

Same thing here, just a little bit different, right? But all in all, guys, that is pretty much it. That is this system. Now, I know I just walked through in depth um the process for the two images. Um just the one image is literally the exact same except the prompt's a bit different and we're not we're only doing one image generation right here. And then for the no image references, it's actually the simplest uh path right here because we do no

Image uh analysises. And it's literally just the AI agent right here, but it's doing the exact same thing. Okay, so the I know it looks very complicated, but all you need to remember is these this path right here is for two images. Same thing for GPT. It's the exact same thing. It is a little bit different just cuz we work with things a bit differently. Um, in some of these requests like for example in GPT, we actually need to upload the

Images to Kai's file upload URL right here. So that's what we're doing here. Um, so just so you know, we're uploading those images there. Okay. And we're all you need to do is just add your Kai API and it will just autoconnect. And then we're that when we're actually making the request for the GPT image, it's a little bit different as well. So this is what it looks like. Um, for the prompts for all them, by the way, we're basically just

Replacing any specific characters that may screw up the JSON here, just so you know, where you have the size here, the file URLs, same sort of thing. We're just referencing it through the Kai uploaded image area. And then you're just the same thing. You're going to need to change out the callback URL here. The variant is just one because we're doing the variance through that loop function that we built um through N8N. And then we're enabling a fallback model just in

Case which for this is just flux max. So it's pretty straightforward, pretty much the same thing as Nano Banana. But yeah guys, like I said, that's pretty much it for this system. We use this system to run ads on Meta for our own AI agency. We use a variety of other different um you know, solutions and strategies of course as well. And I'm going to be making a lot more videos because like I said, this is definitely something that helped us,

You know, attain some of these high weeks and we're scaling very aggressively. So this I know this for some this may be a lot, for others this may be nothing or you know there's every level there's another devil. So um you know you always want more and more when you're scaling your business of course. So, we're scaling very aggressively and I'm excited to uh continue scaling this. And part of the reason why I haven't been uploading as many YouTube videos lately

Is because I've been so focused on the business side of things. And I personally want this YouTube channel to be not just something that's built um and you know, me making a bunch of money just from making YouTube content about how to run your own AI agency when not actually running a successful AI agency myself. Um you know, I actually want to be someone that's actually doing the same thing that I'm going to be telling you guys. say if I'm going

To make a video on how to actually make money with your AI agency that I'm actually running one myself uh day in and day out. So yeah, I wanted to make a very practical video that you guys can take uh get some value from this, start leveraging it um immediately to, you know, make some money. But there are so many other things that go into actually scaling your business, scaling a successful AI agency, being able to deliver really great results for

Clients, systematize your business, hire superstar team members to be a part of your company, and a bunch of other things. So, if you want to see more content surrounding those sort of things, as well as content around NAND, new templates, new practical knowledge, then one, subscribe to this channel, drop a like to show your support if you got some value from this, cuz I did put a lot of time into this template. And like I said, guys, make sure to join

The Stride AI Academy. I'll leave a link down below for the free version of the Stride AI Academy where you'll be able to get the template that I showed in this video as well as the Air Table template right here. Now, I will be adding a few additional bonuses in the Stride AI Academy Pro. So, if you want to join that one, it's the cheapest it's going to be, but I will include some additional prompts, my exact, you know, memes and

Uh styles that you can literally just copy and paste, so you don't have to come up with your own. And if you do want help growing your AI agency with a proven model, you know, you want us to help you with your offer, how to actually sell, take it to market, get a bunch of leads and appointments for your business, book a call on my calendar. I'll leave a link down below where you can speak to myself or someone from my

Team and and we can see if it would actually make sense for us to work together, if we can help you or not. Other than that guys, that's pretty much it for this video. I hope you enjoyed this content. Let me know in the comments down below what type of content you want to see in the future. If you have any questions or thoughts about this video, let me know as well. And then if you want to connect with me, like

I said, join our Stride AI Academy, the free or pro version. And I'm going to be putting a lot of time and effort into those communities um right now as we are just launching them. Once you're in the community, by the way, you can go to the Stride AI automation section right here, and this is where you'll see the video uploaded. You'll see all the NAN videos uploaded right here, where you can get all the different templates and whatnot. Other than

That, guys, I will see you in the next video. Keep hustling, keep grinding, and of course, guys, accelerate your stride. Take care.