Fire your graphic designer, steal this n8n agent instead (free template)

What's up guys, in this video I wanted to show you a simple but crazy effective agent that's going to allow you to essentially fire your graphic designer, or if you don't have a graphic designer, allow you to not have to hire a graphic designer and instead use this agent. It's going to be able to generate logos, ad creatives, landing page mockups, and even assets for your brand. So we're going to jump right in and I'll show you how it all works. And if you stay for the video, you'll be able to see how you can get the free template for this and get started with it straight away. Okay, so we're in NAN now and this is how the agent looks, and I'm going to give you a quick overview of it and then we're going to jump into a demo so you can see exactly how it works. So a really, really simple agent setup. If you've used NAN before, you'll know exactly how this setup looks. But just to give you a breakdown, we have a when chat message is received. This just means that we can chat with the agent via a chat window in NAN like this. And then we have our actual AI agent here, and then we have the OpenAI chat model connected up, which is essentially going to act as our brain. It's the LLM that's powering this agent. So for this, we're using GPT-4.1 mini. And then we have a simple buffer memory. So this just means, if you don't know what this is, it allows the agent to remember, in this case, the last 5 messages in the conversation. So you can change this number anything you like, but right now it's set to 5, so it's going to remember the last 5 messages in the conversation and just makes the conversations flow a little bit more naturally. And then down here we have the actual main tools. So we have Generate Logo, Generate Ad Creative, Generate Asset, and then Landing Page Mockup. And these nodes here are all executing another workflow. So this just means we've made another workflow somewhere else in NAN and we're calling upon it with these nodes here. So I'm going to show you how this all works. Now if we test this out, let's say something to it like I want to generate a logo for my sportswear brand. Then we can send that off and it's going to hit us back with a couple of questions, uh, requirements that it needs to create this logo. So what is the name of the brand? Let's say something like, um, I've got a lamp on my desk, let's just say Lamp Sportswear. An industry or type of sportswear, let's say gym wear. And then what vibe or personality? Let's say fun and creative with an elegant twist. So a little bit vague there. Fun and creative, an elegant twist. You can be as detailed as you like answering all these questions. Of course, if you actually have a brand, you're going to be able to answer these questions much better because you can give more detail on each question. Do you have any preferred core colors for a logo? Let's just say blue and black. And then where will the logo primarily be used? Let's just say on our t-shirts. Okay, cool. So now it's going to use the generate logo tool and it will take a second. It will come back to us with a link and then we can actually view this logo. Okay, cool. So it's come back to us with a link. So if we open this link, we can see the logo that's made for us. Okay, so here we go. Cool little design here. And then obviously got the name of our brand. Perfect. And it's done it with a transparent background by default, which is really nice. So we can actually use this logo straight away. Now we're going to try and generate an ad creative. Let's say I Now I want to create an ad for my brand. And now I want to create an ad for my brand, and it's going to come back to us again with some questions to answer. So what is the name of the product you want to advertise? Let's say Lamp Sportswear, uh, t-shirts. Who's your target audience? Let's say, um, males aged 18 to 30, for example. And then what is a key benefit or message you want to highlight in the ad? Let's say, uh, breathable sportswear tech. So I said breathable, uh, sportswear tech. And what's the platform? Let's say Instagram. What tone do you want for the ad? Uh, let's say fun and playful. And energetic. Do you want to include your brand color palette in the ad? Yes, black and blue. Okay, so left it quite vague, but let's send it off and just see what it comes back with. Okay, so this is what it's come back with. Let's again open this link, a new tab, and you can see, pretty cool, actually looks really good. So stay cool, play hard, added a nice little tagline for us, got our shirt in there, and obviously a realistic looking guy wearing a sportswear shirt. So absolutely perfect. Now let's say, uh, I want to create a landing page in the same style. Then again, it's going to hit us back with some questions for us to answer. So name of the site is going to be same thing, Lamp Sportswear. Then what mood or atmosphere do you want the landing page to have? Let's say fun and energetic. Again, same thing. What should the call to action be? Let's say just buy now. Keep it simple. What kind of feature visual would you like? Uh, realistic action shot of a man in the gym wearing one of our shirts. So kind of similar to that ad there. And then do you prefer a specific grid layout? We're just going to say single column. Perfect. Send that off and then see what it comes back with. Okay, so it's come back with this again. Let's just open this link up. And see what it's given us. Okay, perfect. So we've got a simple landing page design here that of course we could use on our landing page, or we could just chuck to an LLM to actually code us this landing page for us. So perfect. Now we're just going to say, can you create me a quick asset? I want a— I want the name of our brand on a sports drinks bottle, let's just say. So quite simple, the name of the brand on a drinks bottle for the gym. Uh, type of asset you want, let's just say 3D render style. And then the theme or keyword, so for this we're just gonna say bottle. And then dimensions of the asset, uh, let's just say 1 by 1. Let's just have it a square. Uh, any preferred core colors or style hints, super realistic and blue and black. Okay, perfect. Okay, cool. So it's come back to us again. Let's just open this link up and see this asset that I made. Okay, perfect. There we go. Very cool. By the way, guys, if you want to get a free template for this, if you just come over to Communis, you can either just type it up or you can just follow the link in the description of this video. You can just go to templates and you can see on here there's a whole library of all of my templates along with some other creators that have also posted their templates on here. You can type up the one you're looking for, or like I said, you can just follow the link in description and it'll take you directly to it. And then you just hit view details and go, I want this template. And you can see here, you'll be able to download the template, and you can also have this video to go along with it to walk you through in case you forget how any of it works. And then down here as well, there's some more templates that you can also download in case you wanted to. So if you want to grab your free template, come over to Communus or follow the link in the description, and you can get it for completely free. So now we're going to look into each of these tools, and I'll show you exactly how to make them and the breakdown of all of them. Luckily, they're all quite simple and they're all very similar, so shouldn't have too much trouble getting your head around how the tools work. So if we look at this first one, Generate Logo, if we open up this workflow, you can see at the beginning they're all going to look very similar to this. At the beginning we have when executed by another workflow, and if we look in here, the only thing that we're going to be receiving is description. So this main agent is just going to be inputting description into this tool, and it's going to go into a prompt generation LLM. And this is where the magic happens. This is where we're actually going to get specific prompts to use for an AI image generator. So I'm not going to read you through this whole prompt, but basically in here we're telling the LLM that it's logo designer, and we've given it all of these different rules and instructions here on how to create the best looking logo. So declare it as a vector logo and state the brand name. Choose a logo structure, um, icon plus wordmark, emblem, abstract, monogram, etc. Specify style attributes, line weight, symmetry, viewport padding. So all of this different stuff, this LLM is going to be generating a prompt that's going to include all of this different stuff based on what we've given it. And it's going to return just the prompt here. So this LLM is going to output just a prompt that kind of encapsulates everything we've said to it that we want in this logo, and then it's just going to make it way more detailed and way more precise. And then we pass this over to an HTTP request step here, and all we're doing here is we're just sending over an API call to ChatGPT to use its image generation model. So we use this endpoint URL here, and then it's content-type application/json, and then authorization, and then your API key. And then this is the actual body that we're sending over. So the model is gpt-image-1, The prompt is going to be the output from that LLM step, and then you can leave these as you like, or of course you can change the size and stuff. But for us, we're just going to leave it as it is, and these are all going to be square images. Basically what that's going to do is it's going to return to us this massive binary file. And if you don't know what binary is, it's basically just— to us, just looks like a massive bunch of just ones and zeros, random numbers and letters that mean literally nothing to us, but it just contains the data for an image. And what we do is we use this convert to file node here and we set it to move base64 string to file. And then, and then this is the name of the data that gets sent over from the HTTP request. And then we just output it into a data field and this is going to give us an actual file containing that image. And then we just do Google Drive, upload the file. And the reason we do this is so we can get that image link that our LLM or our agent was giving to us. So it's going to give us an actual image link by uploading that image to our Google Drive. And then edit fields, we just grab the web view link from the output of the Google Drive set, because Google Drive will just output a bunch of data to you that you don't need. You can see some of them here, like kind, ID, name, MIME type, all of these different data points here that we just don't need. So we just grab web view link, and then we end the scenario with that. And then whatever it ends with gets sent back to our main agent here through this tool. So if we look at the executions, you can see exactly how this went down. This is the one that we just sent off. So it received brand name, Lamp Sportswear, industry, gym wear, brand vibe, fun and creative with an elegant twist, core colors, blue and black, usage context, on our shirts. Perfect. And then the prompt generation agent outputted this entire prompt here that just went in way more detail and made it way more precise. And then that got passed over to our HTTP request step here that, like you can see, passed through all this binary data that means literally nothing to us. And then we convert that into a file that looks like this. Which again we can't really do anything with, but this means it's actually in a file. The image is in that file. Then we upload that to our Google Drive, get all of this data back, and then we just grab the web view link, edit fields, set the web view link to web view link. So it means it just outputs web view link and that's the only thing that gets sent back to our main agent. So that's how the logo designer one works. If we look back into this, we're going to skip ad creative for now because there's some more steps in there. If we go to generate assets and then view the actual workflow, you can see this one looks very, very similar, and it is the exact same. The only thing in here that's different is the prompt. So if we have a look at the prompt here, again, we're just doing the same thing. We're saying you're a multidisciplinary graphic designer, and then we're just saying what's going to be inputted to it, and then we're just saying again, create a prompt to make this better. So 95 words, it's going to include all of these different things here, basically improve the prompt. And another important thing, by the way, at the bottom, we have to make sure that this is— it's not in JSON, but we have to make sure that because it's being inputted into a JSON body, it doesn't contain any sort of illegal characters. So things like the speech marks, the quote marks, the brackets, that kind of stuff. If you send that over in this HTTP request, you just get thrown back an error because it would mean that this whole thing isn't valid JSON. So this should literally just be in plain text that gets sent over to ChatGPT to generate the image. And then again, we convert that to file, upload it to Google Drive, edit the field to just get the web view link. And then if we look at the landing page mockups workflow, open this workflow, exact same thing as the last two, but the prompt is different. So if we have a look at this prompt, open it up, you can see these are the requirements this time. So this is what is being inputted, and then basically turn that into a better prompt based on this stuff here, which is just style guides and how to improve the prompt. And again, we've left this important bit at the bottom here that gets sent over to ChatGPT, creates the image, returns to us a binary, converts that into a file, uploads that to Google Drive, and then we edit fields to just get the WebView link. So those three, I appreciate, might be a little bit boring because they look exactly the same, but the main difference in there is the prompt. The prompt is what makes this whole thing. And then if we look at the generate ad creative, this is where it gets a little bit more interesting and a bit different. If we have a look in here, You can see again, this looks similar. This is the exact same except we have a switch node in here. And basically this is because for this one you're able to actually upload an image that you want in your ad creatives. Let's just say for the Sports Bear brand, we already had some mockups of our t-shirts or our hoodies, or we already had the physical product and we just took a picture of it. We'd be able to upload the image and then say to the agent, use this image. So if we look at this node here, which starts us all off, we have the description input just like the other ones, but then we also have the image. So when we upload an image, the agent from before can upload the image into this input here, and then if there is an image, it'll go this way. If there isn't, it goes this way. And the way that we decide that is with this switch node here. So if you have a look in here, JSON image, which is just this image input here— if JSON image is equal to no image, then it will go this way. If it's not equal to no image, it'll go this way. And the reason we've done that is because the agent from before has been instructed to say, if there's no image in this input, just write no image. So if the input is no image, it goes this way because there's obviously no image. But if it's anything other than no image, it means it is the actual image that's been uploaded by the agent. So in that way, it will go this way. So if we have a look at this LLM here and we go into the prompt, you can see it's been passed through the description as well as the image, and then also an image description, because when the agent from before passes through the image, it also passed through a description of the image so it knows. And then we've basically just given this a prompt that's doing the exact same thing, generating a better prompt, uh, just like the other workflows are doing. It's generating a better prompt, but of course this time it's including that actual image. And then when that gets sent over, there's going to have specific details on what this image is and then specific instructions to actually include this in the advert. And then of course, same thing, send this off to ChatGPT to generate, convert it into a file, upload to Google Drive, and then edit fields. And then this one down here looks almost identical except we haven't given it the image and the image description because there isn't one. But this prompt is almost the same except we've just left out the bit about including in an image and anything like that. And by the way, if you want to actually call upon another workflow, I didn't explain these nodes initially. It's just going to be a call another workflow node and you just set a name, something that the LLM will be able to use to sort of determine which tool's which. It's just a simple name like generate logo. And then you pick the workflow that you're calling upon. So in this case, it's logo designer. And then down here it will say the inputs for this, which is description. And then if you just click this button here, it's going to be defined automatically by the model. So this just means the agent can just automatically define whatever it thinks should go in here will go into that input. Now we can have a look at the prompt for the actual agent. So if we come into here, this is the prompt. So we've basically said you're an orchestration brain of an AI automation stack. You converse with end users, decide which specialist agent to use, collect any missing information, and pass a clean validated JSON payload to that agent. After the specialist returns its image generation prompt, you deliver the final answer or asset link back to the user. And then basically what we've done is we've just listed out each agent that it has access to here. So, so the ad creative generator, the landing page generator, asset generator, and then also logo generator. And then we also just set the required fields, so things like brand name, industry, brand vibe, and then optional fields are core colors and usage context. But majority of the time it just asks us that anyway. And basically all this agent is instructed to do is when you get an input just simply determine what it's for. So a logo, an asset, a landing page, an ad creative, which is able to do pretty easily. And then just make sure you get these required fields. And then if we scroll down, we just have a little section at the bottom here, sort of like a decision workflow. So classify the user's request into one of the content types above. If ambiguous, ask one concise clarifying question. So basically, if it's not too sure, just confirm with the user. Ask user only for the missing items, grouping them into a single question when possible. And then return to the user the link for the image, and then just the output from this step here. So that is it for the video, guys. I hope you enjoyed it, and I hope you learned something about creating these graphic design agents that, that can allow you to not depend on these people that you've been depending on before, like a graphic designer, or just allow you to not have to hire an expensive graphic designer and instead use something like this. If you did want to work with me so we can figure out the problems in your business and then build custom AI solutions to solve these issues, then, then I've left a link to my website in the description. You go ahead and book free call and we could talk about some ways to integrate AI into your business. And like I mentioned, make sure to grab that free template so you can get started with this straight away. If there's any more videos you'd like to see in the future, please do just let me know and I'll try my best to get around to doing them. But if not, I'll see you in the next video.