I Built an AI That Makes Viral Images & Video in 2 Minutes (And I'm Giving It Away for Free)

A creator shows how to use Nano Banana and VO3 through Key AI to generate viral images and videos in two minutes. The tutorial covers the full workflow and prompt engineering, then offers the tool free to use or resell to businesses.
Guys, I woke up this morning and I thought, you know what? How nice would it be if you could use the nano banana and VO3 model and group them together into one saleable flow that your audience could use and go and sell to businesses for $2,000, $3,000 and I've built it for you. So I'm going to present it in this video and what it's going to do is it's going to take your basic imagery, any client's basic imagery, from looking like this blender to looking like images like this, videos with audio that look like this, in order for you to be able to go and sell them to your clients. Now, I'm going to be consistently producing content like this. You have to stick with me to the end of the video because I need to explain how the flow works. Let's jump straight in. So this is the flow that I'm going to be giving away for free. It's literally going to be in the description of this video. But stick with me till the end because you're going to need to understand how you need to piece this together from your side. We are using key AI and this is a platform that I use regularly. It does have some free credits initially, but in order for you to take it into production from a client perspective, you're going to need them to obviously pay for key AI access. I just find it easier using keys AI documentation because from a key AI documentation point of view, all of the curls for the individual HTTP requests are inside of here. I'm not trying to get technical on this video. Everything will be done for you. It's already there. All you need to do is add your authorization keys to it. But what I'm saying is the reason why I use key AI is because when you debug later, it's just a great tool where you can go to one place and it has support for an abundance of individual AIs Runway alphsedence sunos in here as well, from a voice slash music perspective. And yeah, AI imagery, you can choose from any of these Mid journey imaging for anything that you want. Today, obviously I've used Nano Banana and VO3. I think they are the most powerful that I've seen. You know, to turn images from this to this with just literally an AM form that you can pitch to clients, makes it easy on their side and it makes your job extremely easy because the flow, as you can see, is quite basic by nature, so you're not in a position where you're going to fall over yourself. You literally just have to watch this. So what do we do? Well, the explanation of the flow is quite simple and I'm not going to go into huge detail about it. Then wherever there is an explanation, I've added a sticky note just to give you a basic understanding around what it is. This web hook here, let me just pop on my presentify. This web hook here listens to the completion of this nano banana request here. And it has a callback URL in it. This callback URL will need to be updated to your individual form. So you will need to update this to your form, your production form, in order for it to, for it to work. The rest of it is all written out for you. This is just anybody who's used NAN forms before. If you simply go to your form inside of you, the form that you, that you, you see when, when it's obviously in your NAN instance, you'll see that it'll have a different NAN form URL production URL. You need to paste it in there. Outside of that, everything is literally done for you. Because what we actually use is temp files and temp files is a temporary upload of files we upload to redownload so that the AI agent can understand it. It was what just came to my brain at the moment when I was working on it, about of quickness to put, put this together. But this is production ready. Like you can take this and you can run with it. If there are any edits before I put this video live, then I'll put them in the, in the, in the sticky, sticky notes. But ultimately all we're doing here is setting a field, setting a file URL and downloading, redownloading the same file that we uploaded originally. Now, the reason that there's a set file URL in here is because there's a slight change we're using. This doesn't. And don't, please don't blow, blow your mind with this when you listen to me. If, depending on the level you're at, this might be basic for some of you. Some of you. This might be really technical stuff, but I'm using a replace all function to replace the URL tempfile.org with temp file.org DL which is what I found the download URL to be when I uploaded it literally off the dome. This build downloaded the file and then we have an AI agent. Would you believe it now? Yeah, I hear you cry. You've been saying not to use AI agents. The input to AI agents can accept binary files and text. So that's the reason I use an AI agent in this flow when my brain was mapping it out. And the AI agent here has no tools assigned to it. It's simply just a vessel to pass through information. That's why we use it. Now there is one key thing that I'm also going to give you that is in this flow I have to give it you for you to be able to use it. But something that I urge you to use later on down the road. Because what you might have seen is you might have seen on a regular basis when you're using AI agents. If you are using AI agents, you can have a structured. Sorry, you can have a require specific output format. Okay. And when you come out of there, you'll see the text output parser and you can select JSON. Well, that commonly fails, literally commonly fails all of the time. The JSON output can be incorrect. But if we go and take a look at some of the executions like the one that I've just done, it outputs them in the correct format every single time. Now what that means is that we can guarantee the input that the output that the, that the AI node is going to give by giving it a structured, a structured JSON parser. So it's exactly the same as the built in one. I'll show you quickly now. Or if I got it here. Yeah, I'll show you quickly now. It literally is calls a sub workflow passes the JSON using, using some, some JavaScript that ensures that it's done properly. You'll be provided all of this. You don't need to touch. It literally does it. You can see how it takes the text input that it receives and it inputs it in from adjacent. Outputs it from adjacent perspective in the correct format. So that's all that, that is. But it's a tool that you're going to be able to take and use every time you need a pass. Need to output JSON. Whenever you need JSON outputs, you can use that execution workflow and it will give you a leg up. So there, from there what we do is we split out and we pass the nano banana in the attempt to generate four images. Because four images is the, is the max. From, from a nano banana perspective, four images output. Now I don't force the AI to produce four images if it doesn't think that the output is of high enough quality when it comes down to this, this individual flow. So we will see when we, when we, when we go live and when we look at what we've Produced and I'll run through a live flow review. But looking at the VO3 function, this is, this is all going to be ready, literally ready for you in, in the aspect ratio that we want it, everything that we've got. But what I wanted to try and educate you guys on, it's like you're going to take this flow and you're going to be able to run with it, and that's fantastic. Okay, you've got yourself a flow here. You can sell for thousands of dollars, literally thousands of dollars. Approach your business. Let's do it now. Right, Watch this. I'm going to go to the Facebook ad library, all ads, and I'm going to search for blenders. Now, when we look through these blenders, take a look at some of the. Some of the imagery and video of these blenders that they've. That they've got. Like, this image here, for example, is an example of imagery. Look at the. Look at that. Like, look at that image there and then look at this image here. The quality of that image that we've produced is exactly the same quality of as, as, as of that. But if you. I'm trying to find some examples. They're all videos. I mean, look, you could produce something significantly better than that for Burlinghouse. These images you can create depending on the prompts that you give it. Like, we want images, don't we? Let's see if we can fill up an image media type. Images apply. That's very odd. Let's see if we can see this. I mean, I don't know who's selling using Creatives like that, but that is. Okay, so it's from the marketplace. That would make sense. Look at this here. You can smash that out of the park with some of the imagery that, like the one that I've shown you. There's not. These are marketplace images. So let's just get rid of that. Let's just have one final look. The point I'm making is look at the output of the video quality that I've just shown you and the image quality that, that I've just shown you in comparison to some of the things you're seeing. You see, you could create images exactly like that with using VO3. These images that, that you're seeing here, like this, they're not. They're not of the. That look like these. They're so replicable, it's a joke. And you can improve on what they've. On what they're doing for every business that you look at. And that is Key. So moving away from looking at that, we found a couple of examples of blenders. Not too much that I could find from a Facebook ad library perspective, but you can see you already know from the business perspective. Maybe we go and have a look for blenders. Businesses that advertise blenders online and find some. Let's pick a couple smeg. Let's try these. Look at that image. Like, why would that business not pay you for. For that? I mean, I think that's a blog. But why would the business not pay you for imagery like that? They go for. And look at that. We can produce something significantly better than that for. For smeg. And maybe that's exactly what we try and do here. Why don't we. So let's have a look at the personal blender. Let's take the smeg blender. Take this image. I'm going to save this image. I don't want your promotions. I've already bought your kettle. Smeg blender. Right. So now we're going to use Smeg Blender. And I'm really going to dig deep in here into this because the flow is free and connecting them up is so. Is so easy. All you've got to do is go to key AI, create an account, go to Settings. I'm on the wrong page here. Or not. No, we like it dark. Let's go to the dashboard settings in key. AI API key. And your API key will show create as a credential inside of the. Inside of the notes inside of here, where we make the HTTP requests for nano, banana and VO3, both of these. And you're cooking. You're literally aware you'll have to connect your own models, obviously. Here, Drive. And appending a row. Google Sheet. Literally by id. You still put your ID in there. That will work. Google Drive. I chose that by drive name test creatives. But you can. You can even put the ID and do whatever you want. The point is we're gonna create creatives that are better than. Better than what they produce. Now I'm gonna double up in this video because I'm gonna have AI Studio create me a prompt for this. So AI Studio, turn the temperature down. I'm using Gemini 2.5 Pro. And one of the big things I've been mentioning lately is, is the emperors versus the operators. Right? And what to be. To be an emperor, to be a builder. You can't just type in here. Now please give me a system prompt for this. Right, so you're gonna have to stick with me whilst I. Whilst I Generate this. So what I'm actually going to do is I'm going to add a system instruction. I'm going to write the system instruction myself. Okay. And what I'm going to do is I'm going to play into two of the big boys in the space. You are design that you are tasked with using Nate Herc and Nick Sarayev to have a discussion about prompt engineering. A basic text prompt for video and text output using VO3 and nano banana. They should, they will be given an image and an explanation and they should argue best practices via dialogue. And then at the end produce a. A prompt that outputs, I don't know, four images and one video, but written in plain text. Kind of over engineering this up here. We're not over engineering it, but that's not the best prompt that you would ever give it. But I'm writing and I'm on time to keep you guys engaged. So we'll see what this is. So now we're going to see. So now I'm going to say let me go and get the. The SMEG image that I had. This is attached. Is a personal blender from smeg. I need the best system prompt to produce a lifestyle image that beats the brand zone. You should be extremely specific with your prompt arguing best practice and set the whole scene for every image, every image. And the video from second one to second A should be driven by your prompt and let's hit run and let's see if we can beat it using what I've just produced. The core key understanding around this flow is that SMEG are a massive company, right? And they're creative. Even they're creative. Going back to where we were, even their creative does not beat what we can produce. Right? It doesn't. So straight away you can see how one of the biggest players in that space, high end household items, if you like. I have their products in my house, the high end items, if you like. We can actually improve their creative. So if we can do that, imagine what we can do when it comes down to SMBs that have, you know, one in house marketing guy or are trying to go and type prompts into AI every single time this, this, this happens where they fall into the operator category, which is exactly where you don't want to be. So there's a discussion that happens. Look at this. A SMEG personal blender. Clean, retro, minimalistic. The runzone shots are just sterile products on a white background. We're not selling an appliance. We're selling a feeling of a perfect morning. We need a narrative so they have a discussion. This is part of the prompt engineering around what I've got. And here's the final prompt. So output Falstill images style photorealistic aesthetic. Let's just see what this, what this comes up with because that doesn't look hugely fantastic to me. Maybe I'm going to ask it, please, can you rewrite the system prompt? System prompts? No, we're not going to. Do you know why? Because we're going to test what we actually just seen. We just want to. The reason I'm cutting out the output is we already know what the output is. We already know. We'll take the four images in one video from the output perspective and we will go into here, we'll hit execute on form submission, we will put that in and we will choose a file and we will select the smeg blender and we'll submit. And now we'll see the quality of the creative that this comes out with. And ultimately this is the test. I've used a prompt that's come from the. The AI studio. I hope you understood what I was doing there, that you need to evolve your system prompts. You can ask AI to produce your system prompts that you then tailor. But what I've done is instead of just asking it, hey, write me a prompt. I've tasked Nick Herc and Nick Sarayev to have a discussion of prompt engineering. One of the things I missed is this. They are hugely experienced prompt engineers. Engineers and no nan like the back of their hand. Everything from system prompting. Extremely detailed tasks. Extremely detailed, extremely in depth. From selection workflows to writing JavaScript, everything from system. Everything from system prompts to. To writing JavaScript. They are best in class at let's say that. Right? So we challenged Nate and Nick to create as a prompt. It's now time to check out the results of said video and imagery which is going to be interesting. So let's go and take a little look through here. We have the first image which looks like a close up, which looks absolutely fine. Nothing wrong with that at all. Not exactly fitting on the page of what they want. But that is perfect for like a carousel or a collection. Advertisement. We have another image here. I think that would be better if it was actually blending at the real time. The prompt missed that. But excellent quality imagery. Beats their quality of imagery by by far. And then the big one, let's check, let's check the video out. Let's check the video out and see. I mean that is absolutely fantastic. When you're talking about imagery and image generation and video generation for E commerce businesses. In this case, we've produced better content than what they can produce internally, which is scary because they're not a small firm. So that means that we're in a position where we can be confident that the product that we have and take to market is a. Of sub. Is of substance, is of really key and core importance and can actually deliver better than what the, deliver better than what the, the, the, the major businesses can. And if we can do that, then every single SMB you speak to would snatch your hand off something like that. If you would, you could, you can replicate that flow for everybody. Create them their own key AI. That's all they need, their own key AI and then use that to be able to resell. Every single business is going to be able to use this flow that I'm giving you using key AI. As long as they produce their own, they create their own account, they give you access or whatever. They, you get their API key, you update the API keys inside of the flow and from there you, you, you have a resaleable flow that you can go across any business with or even your own. One final time just to run through this. Let me, let me hit refresh. Let me hit refresh. So one final time we upload the file, we re download, we pass it through an AI tool with a system prompt. Feel free to try and make this more concrete. If you wanted to update it, you can play test with it. That's what how I name them. Like if I'm messing around with something like now and I'm playing with it, but I have the original so I can, I can do this. You also have obviously version history. The safe JSON parser is something which is a bit of a huge bonus in this video that is something that you haven't, you're not going to have had previously, but will save you a significant amount of headaches when it comes to the output, especially when using AI assistant. So if you're ever using an AI assistant, always use this safe JSON out output parser. It will ensure strictly the JSON is exactly how you want it to be. Make sure that you're using that. The, the nodes that you need to update fall into these categories here. So inside of the individual node for Nano banana and for VO3, you already have the individual curls imported from Key AI so you don't have to worry about the structure of the nodes. But what you have to have, what you have to have concern with is The URL that is inside of the node that matches your your URL. I will edit this and type it out so that you you know to insert it there. You also inside of there need to need to add your credential. My key AI is here. You can pass them as headers as you've seen me do historically as well. Like you can pass them as a header. We're using a header off, it's a bearer bearer off approach token. But in this case I save key AI because I use it a lot. So I wanted to give you the that sort of approach and you can't see my API key as well. It makes it safer. So you just need to replace this with your own individual header header AI or add send headers authorization authorization and then bearer bearer and then go to key AI, copy your AI key and paste it in. So it's really as simple as that. You've got a plug and play, you've got a plug and play flow here where you can go and sell this to businesses. I really hope you appreciated it. If you appreciated the video, please like subscribe, leave a comment, let me know that you enjoyed it. If you want to see more of this type of content, I need to know. So you need to comment and tell me this flow is going to make you some money. Those are the ones that I prefer to produce. Educational ones are important too to get the mindset right. But these ones, when it comes down to actually building flows out and providing value, I think it's critical that I'm in a position where I share this with you. You are now leveraging industry leading tools to produce industry leading content for businesses at the drop of a hat that you can duplicate and go and take on and sell to other businesses at scale. And that is something that you don't see consistently. So like subscribe, comment and I'll catch you next time. Cheers guys. Bye.










