This Veo3 Prompt Factory automates $100,000 Ads for ANY Brand (& in Vertical) on n8n - here's how 🥚

[Music] These VO3 ads are going viral on Twitter right now and probably on every platform pretty soon. And for good reason,

pretty soon. And for good reason, because they're seamlessly making marketing ads that used to cost hundreds of thousands of dollars to make, but this time they're being created for literally less than a dollar using AI.

In this video, you'll learn how to automate these golden prompts, which I organize here in this document with proper attribution to their creators, of course, and which I'm sharing as reference for you, so that you can instantly customize them for any brand

just by typing a brief for your AI agent here, and even generate videos in vertical if you wish to create ready to publish marketing ads like this IKEA one from Salma, this Tesla ad by Michael,

this sneaker launch ad by Heather, this Jeep ad by Gilmo, or this Remoa ad also from Salma, and several more prompts and templates in this prompt library, which I organized so that you can just copy them and use them immediately. And if in

case this looks intimidating, don't worry because this tool is actually all no code. And in fact, the actual

no code. And in fact, the actual workflow is just this one. So it's much simpler than it looks. I just included the whole prompt library here so that you can swap out those prompts as well depending on your use case. So if you want to learn how to do this and more, then invest some time to watch till the

end and I'll teach you the principles of how to set up something like this and also so that you can stay across on the bleeding edge of what's possible today with AI. Let's begin.

with AI. Let's begin.

By the way, if you're new here, my name is Jay. I lead Robolabs which is our

is Jay. I lead Robolabs which is our agency arm helping mostly brands and content businesses use AI in meaningful ways and also the Robbo Nuggets community our education arm where we have hundreds of AI practitioners and professionals located across the globe.

And here at RoboNuggets our mission is quite simple is to make AI easy to learn for anyone who wants to so that you too can take advantage of this historic shift that AI is bringing. And in fact to make this as easy as possible I've collected all of those amazing prompts

in this library. And also I'll be uploading this AI agent template to our community so that you can just import it into N810 which is the no code tool we're using if you're already wellversed with this tool and if you want the shortcut but if you are a beginner don't worry because I will be walking you

through this AI agent workflow node by node so that you too can understand the principles of how to make an automation like this which is what's more important anyway in my opinion. All right, so broadly for context, I first saw this trend via this tweet by Men Choy, who is

a great follow in the AI space if you're not following him yet. But I saw this V3 IKEA ad, which now has 2 million views just a few days after posting it, where his prompt was originally inspired by Salma, who seemed to have started this

trend. But now these V3 ads are all over

trend. But now these V3 ads are all over Twitter and probably on other social platforms as well pretty soon. And for

good reason, because these prompts seem to have unlocked completely new capabilities of V3 that we were not even aware of when it first launched. And I

think what's unique about these prompts once you go to this library is that the prompts themselves are actually formatted in markup language. So this

one is in the YAML format and then if you scroll down this one is in JSON which stands for JavaScript object notation. So whenever you see these

notation. So whenever you see these brackets in braces that's basically what JSON is. And I guess it makes sense that

JSON is. And I guess it makes sense that these prompts are actually providing higher quality output for V3 because for years now a lot of the internet and essentially the computers powering it have been communicating through formats

like these behind the scenes. So if you provide a prompt as structured as this into an AI model like V3 which was trained from the entirety of the internet then it makes sense that these prompts are also generating much higher quality output for us. So I think that's

the big unlock that this trend has revealed to the world. But now obviously the question becomes if you have access to this prompt library as well as these detailed prompts how do you actually use them? Well first I'll show you how to

them? Well first I'll show you how to use them manually meaning copying and pasting it into a tool and you get an output in return. But also I'll show you how to do this automatically so that you don't have to tweak those prompt yourself for the brand that you want to

create for. So, first the manual way. We

create for. So, first the manual way. We

basically just want to copy this prompt and put it into a tool that has access to V3 Tree. A lot of people use the official Google Flow tool or even Crea AI, which you can do as well, but these tools actually require a subscription.

So, the cost barrier to entry is much higher if you just want to test things out. So, the one that I found that's

out. So, the one that I found that's also the cheapest implementation, at least right now, is Key.AI. And what's

good about it is not only does it let you use V3 manually, it also lets you use V3 in an automated manner thanks to their APIs, which you'll see in a bit.

So, if you sign up there and head to your playground under V3, immediately you'll see here an interface which you can use to just paste in your prompt like what I did here. And what's great is they actually provide you an option for vertical format. So, let's do that

to just illustrate it. And then the last toggle you can use here is either V3 fast, which is the cheaper version. This

costs around 55 cents for a vertical video and 40 cents for a horizontal video. So, this is in key AI credits.

video. So, this is in key AI credits.

So, you can just do the math when you explore this website. But of course you also have V3 quality which is much more expensive but that's available for you if in case your group has the budget and also V3 fast quality is not yet enough

for your use case but for us V3 fast is good just to illustrate and you'll be able to generate that here will take a few minutes and there you go it just generated this ad which looks pretty good and obviously you can download that as well manually here. So now the

question becomes how do you customize those prompts for the brand that you want and how to do that automatically so that you don't have to write the prompt yourself. Well, the tool of choice that

yourself. Well, the tool of choice that I feature a lot in this channel is called N8N, which just in case you haven't heard of N8N, it's a no code automation tool similar to Zapier or Make.com. But very clearly, it's gotten

Make.com. But very clearly, it's gotten popular because of how easy it is to implement AI agents for this case. And

it's also not that complex if you have someone guiding you through it. But if

you head to N8 through the link below, you'll be able to sign in and try it out and you'll be able to create AI agent workflows like this one that we have here. So, this may look complex, but

here. So, this may look complex, but honestly, the way that it works is very simple. Whenever you run this workflow,

simple. Whenever you run this workflow, you just open this set direction node and in here is basically where you provide the direction to your AI agent on what to do. So you have a creative brief here and you can go as detailed as

you want. The prompt inspiration which

you want. The prompt inspiration which you can just copy paste coming from the prompt library. So you can see is the

prompt library. So you can see is the same thing here. And then I've made it so that you can also define the aspect ratio here as well as the model which we've set as V3 fast and vertical for this case. So that's the briefing that

this case. So that's the briefing that you just need to provide to your AI agent once you have this automation fully set up in your workspace. And if

you execute it, it will basically provide that direction to your AI agent who will recraft that prompt and then pass it along to V3 which then generates the video and puts that video to your Google Sheets for proper logging. And if

you look at that Google Sheets, it was able to generate that video for us which if we click it will take us to a new browser tab.

[Music] And there you go. Our AI agent just recrafted that prompt and made an ad for Volkswagen in this case. So let's go through this workflow in more detail so that you can understand the principles of how I set it up. And the first

principle is this workflow. You can see the trigger for it is on click which is us clicking the execute workflow because I actually imagine this N8N template as something that doesn't run on a scheduled banner. It's sort of like a

scheduled banner. It's sort of like a mini application for V3 generations. So

whenever you need an ad, you can just go here in your mini app and run these notes, which is important because if you're spending money for creating the ads, you want to make sure that this AI agent is actually providing you with the right prompts. Of course, you can change

right prompts. Of course, you can change this trigger to a schedule if you want to generate them every day, for example.

That's quite easy to do in any. But at

least for most use cases, I think it will function as sort of a standalone mini application. But broadly, you can

mini application. But broadly, you can see there are just four steps here. The

first one which we just walked through is this node that will provide the creative brief to the AI agent. Next,

we'll look into the config of how this AI agent was set up in the first place.

And then with its out, which is going to be the refine prompt, it will pass that along to V3 still via the key.ai integration to create that ad. And then

once we get that ad, we just log it to Google Sheets. So I'll just talk through

Google Sheets. So I'll just talk through the setup for this and make it as beginner friendly as possible. But in

the interest of time and for those who are already intermediate level in N8 who are probably the ones watching this, feel free to check out some of our earlier videos where I show how to create workflows in N8 from scratch so that you can better learn and understand this setup if it becomes too complex.

But in any case, starting from top to bottom, the first step is basically this node that's named as set direction. It's

an edit fields node. So if I open that up, that's basically what we saw earlier. So I imagine these fields to be

earlier. So I imagine these fields to be pre-populated by the user every time this runs. So if I just click execute

this runs. So if I just click execute this step, what that just provided as an output is all of those fields that we've manually set and load them into any end so that we can pass them along to the succeeding nodes. So that's all step one

succeeding nodes. So that's all step one really is. But just a note, you can see

really is. But just a note, you can see there's a lot of other edit field nodes here and I just included this here for quality of life. So they're the same ones that we have in the prompt library in that notion document. But if you open this up, this just has the prompt that's

provided by this user that is pretty much ready to go. So that you can just swap this node with any of these and that would hopefully make it much easier to work with this template. But now that we have a creative brief, now we just pass that along to our AI agent. You'll

see that for every AI agent in N8N or otherwise, you usually just need to provide a user prompt, which is sort of what you type into chat GPT, which is provided here. And again, this looks

provided here. And again, this looks complex, but if we expand that, you'll see you have some green text expressions here. But in reality, how that

here. But in reality, how that translates is it's just pulling from your input, which is the brief in this case. I want an ad for Volkswagen. So

case. I want an ad for Volkswagen. So

that will be what will be passed to the agent. And we're also explicitly asking

agent. And we're also explicitly asking it to take inspiration from the prompt that we provided it, which is this long one that we got from Twitter. So that's

good that you've provided a user prompt, but AI agents that are robust also would need a system prompt. And a system prompt is like a declaration of an agent's role in its life. So if you open that, we already have a pre-dealed system prompt here which you can just

read through in your own time. It's also

using our agent framework which we talked about in the previous video. But

essentially let's just make sure that our agent provides the output that we need. So if you execute this step, what

need. So if you execute this step, what is now doing is thinking about how to tweak this prompt that we just provided it for a Volkswagen ad. What it provided us as an output is this title just for logging purposes, but more importantly

this final prompt which we're about to send to V3. You can see that this AI agent has access to these tools or models. So this probably familiar to you

models. So this probably familiar to you already if you have worked with NADN.

But essentially to just run through them, the most important piece here is this model which in a way functions as it brain. So in this case we use

it brain. So in this case we use OpenAI's models which if we open that we're using 4.1 but there are several options here if you want to play around with that. But more importantly if

with that. But more importantly if you're setting this up for the first time you just need to set up your credential here by creating a new credential. And you can see it's going

credential. And you can see it's going to ask for your API key which you can get just by going to this URL and creating a key there which you can paste here. So once you have that set up

here. So once you have that set up you'll be able to access chat GPT from within N8N. Now this think tool is just

within N8N. Now this think tool is just a good note to add as best practice.

This is unique in N8N but basically what it does is it allows your AI agent to reason or think about its output before providing it to you. So I didn't even change the default description that N8N has here and it's simple to add. So it's

just good to have it there almost always. And then we have this structure

always. And then we have this structure which is going to be an output parser node. And what this basically does is it

node. And what this basically does is it just provides the output structure that we want our AI agent to adhere to. So if

you remember the output that our agent provided is an attribute called title as well as the final prompt. The reason why it was able to do that properly is because we provided it this strict guidance that that is the output that we

want. So this is where you declare that.

want. So this is where you declare that.

If you want to add more elements here like a caption and such, this is where you do it. And then finally, we have this Google Sheets node, which is also a tool. But if we open that, that

tool. But if we open that, that basically is connected to our Google Sheets account, which again, if you're new, you need to just set up your credential here. This is simple. You

credential here. This is simple. You

have a single sign on with Google. And

what this basically is doing is it's connected to one of our Google Sheets, which is this template in my case, which I'll also share along with this whole N8 template. And if you remember, the

template. And if you remember, the format of that Google sheet is quite simple. We're just logging everything

simple. We're just logging everything here so that we have proper tracking.

And you can see here that we're specifically looking for data in column B, which is this one. So now if I execute this step, what that basically did is provide our AI agent with ideas

or topics that it has already generated in the past. So let's say you have one brand that you want to generate this for and you want to generate multiple. This

is just a good way to help prevent your agent from duplicating on the ideas. So

that's an optional technique if you want it, but you can also just remove this tool if in case it's not needed for your use case. So there, now that our AI

use case. So there, now that our AI agent has refined the prompt, how do we then pass that along to V3? Well, if we go to step three, there is this node called an HTTP request node, which if

you haven't used these before, an HTTP request is one of the most used nodes in N8N because it's the primary way by which we call on third party tools to use in N8N. So this one node alone makes

N8N very flexible to use. And because

we're making a request, we have the method set as post. So it's like you're posting a mail. And with that analogy, if you're posting a mail, you obviously would need an address on the other end to send that mail to. And so this URL is

what it is. You can see it is pointing to key.ai. So this is just their

to key.ai. So this is just their official URL for automated requests like this. And then obviously if you have an

this. And then obviously if you have an address to send, you also would need some sort of message in that mail to inform key of what you're requesting for in the first place. So if you go down, this body area is where you declare

that. And you can see the body is again

that. And you can see the body is again in JavaScript object notation. So if we open that you'll see that this is quite a simple expression where we are providing it the prompt we're providing it the model and providing it the aspect

ratio all green text expressions because we want them to be dynamic values in reality how that will translate is we're just providing it the prompt we're just providing it the model as well as the aspect ratio which we predeclared when

we first set the direction in the very first step. So this JSON format template

first step. So this JSON format template is just as per KI's documentation. So

I've read that and worked through how to connect to their application so that you don't have to. This is basically how. So

now if we go back continuing the analogy the last thing that you will need is to set up your credential which sort of acts like a stamp to declare to key.ai that you actually paid for the mail that you are sending to them. And there are multiple ways to do this. The setup that

I recommend is to have the authentication set as header O similar to this. And then for the credential

to this. And then for the credential itself you just want to click here create a new credential. Name this with whatever that you want. So sample key AI credential. The name needs to be typed

credential. The name needs to be typed as authorization, taking note of the capital letters and the lack of spaces afterward because if you mess up just one character here, it actually won't work. It's just as per key's rules. And

work. It's just as per key's rules. And

then for the value, you want to change this to expression just so that you can see what you are typing and type in bearer space. And again, don't ask me

bearer space. And again, don't ask me why that is. That is basically how the internet has functioned for these JSON prompts and APIs which are application programming interface for quite a few

years now. But anyway, you will need to

years now. But anyway, you will need to provide your API key here from key AI.

So, how do you get that? So, well, if you go back to key AI, you can see that there is an API key section here, which you can click. And here is where you can create your API key that you can just copy. Do note that your API key should

copy. Do note that your API key should be private because if you share it around to friends, then they'll be able to access your account and also drain your credits and stuff. So, just take note of that. So, if you paste it here and click on save, then you'll have

every element that you need to connect to KAI. And every time this node runs,

to KAI. And every time this node runs, so let's click execute step, it will be able to connect to key as well as V3.

And if you receive a status code of 200, that's the best code of all when it comes to APIs and it will say success.

Now this success means that you have successfully posted your request. It

doesn't mean that the video has already finished processing, which is why this next node is just going to be a wait node. So if you open that up, in

node. So if you open that up, in reality, you can set this as something like 600 seconds for 10 minutes to give it enough time to generate. But for now since you are testing this out, you can just put in 1 second and execute step so

that we can move on to the succeeding step. Now if you go back this next node

step. Now if you go back this next node is just for us to get the v3 generation that we requested from keyi and so if you open that this time the method is just a request to get and then for the

URL itself it's again pointed to key specifically this URL where according to their documentation is where we get our videos. And then since you have set up

videos. And then since you have set up your credentials, you'll be able to choose that as well here. And then for key to return to us the actual video that we generated, we'll just provide it this task ID which is a dynamic value

coming from this previous node. So this

is the ID that got generated for us when we made the request which just lets a service like KI look for the video in their database somewhere. So if we execute this step, it will probably not provide us with the video yet. So if we

go to schema, you'll see that the response is still null because not enough time has passed for the video to be fully generated. So we still have to wait a bit for that. So for you to see where that is, you can just go to logs

here, go to V3, and then here you'll see that it is still running. So probably

grab a coffee or something and wait for it to finish. But if you refresh this and go to v3, you'll see that now when that is successful, you can just go back to your nn instance once again,

re-execute this get v3 step. And this

time we should get a response back as well as this result URL which you can actually copy and paste to any browser tab. And there you go. It just generated

tab. And there you go. It just generated that new Volkswagen ad for us. So that

sliding piece is actually pretty cool.

So now if we go back, the final step really is just for us to log this ad in our Google Sheets template. So if you open that, that just has the credentials set up for your Google Sheets already if you've already connected. This time the

action we're doing is to append a row.

So we're just adding this row in that sheet. And once you have your Google

sheet. And once you have your Google sheet connected here, the columns here should autopop populate. If we open our Google sheet here, you'll see that it is basically the same thing. For the ID, we just have an Excel formula here which

translates in practice in Excel into a proper ID that increments by one so that we have easier ID tracking. Here you

have the title which is important because the prompts here are actually quite descriptive so it's hard to understand what they are without a shorter hand title. And then of course you have the prompt here. We have the production status set as done and the

final output which is the most important because that is where the video resides.

So if we execute this step, it should add another row here and also provide us with a URL there at the end. And there

you go. Step by step, that's how this N8N mini app V3 ads factory works. In

practice, the way you work with this is you set the direction here. You click on play and then you click on play again on the AI agent so that you can check its output before passing it on to V3, especially if you're trying to save on

costs. And then of course, make sure

costs. And then of course, make sure that your wait node is set to a long enough time. So 10 minutes is usually

enough time. So 10 minutes is usually enough. And there you have it. I think

enough. And there you have it. I think

some final thoughts. One is what's cool about this trend is we're actually exploring a bit more about the capabilities of V3 which is quite unique in a new piece of technology because previously if you have software that you

launched it's actually very deterministic. So you can pretty much

deterministic. So you can pretty much predict how that software will behave and be used. So websites for example when they began a programmer somewhere coded the back end to make sure that you can navigate however they want you to

navigate. But with technologies like V3

navigate. But with technologies like V3 and other AI models, even though they were released already months ago, we keep discovering new things on how to use them, which is just the nature of AI, which to me is really interesting and which this trend has revealed. And

then a second thought in terms of a usability perspective is that some operational questions come up, right? So

we want to be able to create an ad not just for brands that V3 is probably trained on. So if you're thinking to

trained on. So if you're thinking to launch your own ecom brand, how do you then feed that to V3 as a model and create ads from that? So I think there are ways to do that now. probably not

yet perfectly. So, if you're keen to see us explore that, then stay tuned for more content and subscribe to the channel. And then a second thing is

channel. And then a second thing is right now V3 is limited to 8-second videos. How do you make it so that it's

videos. How do you make it so that it's a seamless 30-second, 60-cond ad instead of just these 8-second shorts? So,

that's something I'm excited to touch on for succeeding lessons. So, stay tuned for that one. And also, if you want to learn more of how to set up an automation template like this, feel free to check out the RoboNuggets community, which also functions as our education

arm. We have tons more lessons here

arm. We have tons more lessons here around artificial intelligence and automation as well. Several of our AI practitioners around the globe are also posting paid opportunities and partnership collaborations here. And we

also host learning sessions with well-known tools in the AI space and even do meetups IRL occasionally, which is pretty cool to see. So, if you want to join a community of AI practitioners to navigate this quickly evolving space,

then have a look at the Robbo Nuggets community and see if that's for you. All

right, guys. That's it for this one. See

you guys next time. Thank you.

[Music]