How to Make Al Video Ads With One Prompt [Full Workflow] | MiniMax Design

So this ad was not filmed, edited, or scored by a person. And it came from one sentence typed into an AI agent. And that means no prompt engineering needed, no editing timeline, and no separate audio tool. The agent planned the shots, generated the video with sound natively baked in, cut it together, and then added subtitles. And so in this video, I'm going to show you step by step how to do exactly that. And we're going to create a set of assets for a fictional coffee brand. That'll be a UPC ad, a cinematic kind of motion graphics piece, a whole bunch of stuff. I'll show you where it falls short and then how to get the best output out of a system like this. That's going to be a video. It's going to be a fun one. Let's go. So here we are at minimax.io. We can see all the openweight models and the product side of it on the screen here. So let's jump into that minimax design platform. Once you get to this page, you'll go ahead and download for your system. And now this is what you'll see when you download minimax design for the first time. It's an interface that you're going to be very familiar with at this point if you watch the channel. And we can start to see some of the canvas-like background here, which is how we're going to engage with our design agent. Now you can choose the models that you want, like your newer GPT-6 Astra models. And so the minimax design platform does not only use their own models, they rather optimize cost and performance for you across the Pareto frontier of different model cost and performances. Their models, for example, are really good while being extremely cheap. But if you do want a little bit more performance, then maybe you go for one of the other models you can see here. And generally, you can see this, in fact, where it says speed, smart and cost. And then each of the different models is going to have its own kind of comparison. So let's just jump straight into it. I'm going to create a 15-second UGC-style product review for an Alto cold brew coffee brand. And then I'm describing what I want that to look like. I'm saying I want one person speaking to the camera about the taste and the caffeine, and we want it to be vertical for TikTok. So we're not playing with any settings. We're just telling the agent, hey, this is what I want. And then we're also going to add some references. So right here, we can see our different logos. This was the final logo we went with. So I'm going to drag this into our prompt. Then I will reference this image right here using the at key. And I'll say this is the logo or the brand. And let's let that run. And so here we can start to see the beauty of the canvas interface, right? So if you're familiar with Figma, that's a interface that you may be familiar with as a designer. And so anywhere we go around here, if we select those pluses, we can create a new element. Also something you'd be very familiar with if you've worked in a system like Figma. I won't go too deep into this, but effectively, there's some organizational tools up here. If you click the mini map, kind of like a video game, you can see where you are on the map or the canvas, so to speak. And then your actual editing tools, if you press on the image, you're going to see here. And there are a lot of tools there. Redraw, Relight, Storyboard, just think of that as kind of your vertically oriented Photoshop tools, just flipped horizontal and added right here. And then on top of this, you have the Gentic capabilities. So you want changes made to something, I can just press on the image. So AI is pretty much built into all of these editing tools right here. I can also hit the at, I can select the image and say, for example, make this background green. If I wanted to level up even more from there, I could say, make 15 variations of this, each of which has its own background color, so I can see which one I like the most, right? So there's a kind of scale to the creative potential here, not only with this form factor of the canvas, but the additional layer of AI. I think this is where all design tools and platforms are going. Your Adobe Creative Suite right now is playing a catch-up game. Platforms like this, I think are actually much more on the bleeding edge of how people are creating a lot of graphic and design assets. And you may say to that, well, there is a vector editing capability that you need if you're at a certain kind of tier of working with graphics. And there is AI for that too. There are models that create in vector form and let you edit like so, or make changes, have AI make changes on a vector design, on a vectorized design. Okay, now the AI is asking us questions about our 15-second TikTok video. Let's see, let's do the unbox first. I'll click and let's use a generic logo can. I'll click submit, it's going to keep working. Okay, now it's asking us what kind of creator we want to be in our ad. I'll just go with what the agent recommends for us. And then I want to categorize this under a project while it's finishing it up. So let's say local project, we can make it cloud-based or collaborative as well. I'll say UGC ads for the brand we're working with. And then I'm going to move this project into UGC ads, right? So I could go and I could have five, 10, 15 different projects or do it all on one canvas where I'm making the ads for my business. I'm also going to create right now one more project that'll be or something we're going to jump into after this, which gives you much more control over what you're making and is a pretty cool feature. So that's coming up next. Okay, so now it's generated an image of a creator for us and it said, hey, does this work? Is this what you want? Very, very collaborative as an AI. This looks good to me. Very good image actually, but let's just say use this creator. We'll click submit and then now it'll proceed into the video stage. We can also see the storyboard that was generated here specifically. So we have our actual second by second breakdown of the shot. So that's pretty cool. And our creative kind of take as well and what it's visually going to look like. This is less for me and more so for the AI because this is how it proceeds, right? It builds on itself step-by-step to get to the final asset. And then here's where we can see the video that's being generated and we can see visually how it connects to the assets that were created prior. So our logo and our model right here. Okay, here's our video and we can see off the bat that the references held up extremely well, right? So we can see our reference logo is very high quality on the can and the person looks realistic as well. Let's give this a watch. Unboxing this for my afternoon coffee. The Alto logo is simple. The can feels clean and okay, this is really smooth. I like the clean caffeine feel. Coffee flavor without that heavy bitter finish. I'd keep this in my fridge. You can see exactly how good that is, right? Very smooth, very realistic. The voice has kind of the tonality. I think with the okay, that's kind of the highlight moment for me. You know, we did it in five minutes, right? So whether you want to pull end-to-end, create these UBC ads with AI, or you want to just do scenes in them and have your own narration or vice versa, whatever it is, this was one-shotted in 10 minutes of work. And if you wanted to sit down and make 20, 30, 40 of these in a day, explore different concepts, that is the value of AI for design work right here in this video. And that's why I'm such a proponent of it for you in your business and whatever form that looks like for you being able to do better work and to do it faster and cheaper. What we just saw is very high quality work by the AI will save you a lot of time. But at the end of the day, if you watch this channel, you've seen a system like that before. What you have not seen is this that's coming up and what minimax design has done is built out a kind of node-based architecture that gives you much more granular control and kind of using AI and an AI native system, but building into your creation. So we can click this big plus right here, and we can see these different nodes, right? We have our text, our table, image, video, audio, those are your basic set of elements. For example, if I click table, right, we can see a table, text is a text element, all once more, very much AI native, but then we can get into something like a 3D director stage. So what we're going to do is create another video, like an ad for our brand, but we're going to do it in a fundamentally different way that gives us, again, more control over the architecture. So I just clicked plus, we can see right here, uh, the form that is opened and let's click open director. And you can already start to kind of tease out and see why this is so valuable because what this does is it gives you literal shot level 3D control over the shot that you're making. So we have our character, we'll have our background in our scene, and we're going to build into the video that we're creating with AI. So let's build this shot by just talking to the AI. So let's say, what kind of shot should we build for an ad for our coffee brand called Alto? Now there's a lot of depth in here, we're not going to get through all of it. What I highly recommend is if something like this interests you, download this, spend 30, 45 minutes playing around in it to understand all the tools, because this is something we haven't really seen before on the channel. It's something that takes a little bit more time to get to know and properly utilize as a user. Okay, here's the shot it recommends, 16 by 9, and three quarter dolly in at tabletop. So we can go to add a path, and then we can do a dolly in right here. And then now we can see on the timeline here, so we have this four second shot where we have the dolly that moves in. Now let's jump into characters. Let's just click AI generate and say we want a coffee shop worker, we'll leave it open ended, that will design our character. We can also see the actual prompts we may want. Now keep in mind when this is generated, right, it'll fit in with the rest of the scene, that's the job of the AI. This is giving us just more directive, right, creative director control over things like the camera movement, where objects are located, etc. As opposed to necessarily the visual design of said assets, which we can get by just talking to the stage agent. So just for fun to demonstrate here, let's go ahead and put a bicycle on the ground here. So we're going to do that just so we can see for sake of demo in the final video, what it looks like to have that asset place. For this we can go to Google, we can say outdoor coffee space. Something like this photo from Instagram works, pretty low res, but it'll work for us. Gonna click upload. Okay, here is the final video. After some more work, let's give this a watch. So there we go, we can see the bike, the two people, the same background that was used, and our reference image right here. So that's exactly what we wanted, and you have that kind of control now over whatever kind of ads you want to make, if you really want to go the extra mile and have that vision to pull from in mind. Okay, now for demo number three, I want to jump into motion graphics. That's its own whole world, and this platform is great for that too. So I want to show you how you can create motion graphics and text overlays on screen with this system. Let's click create, then let's click new creation page. I'll just call it motion graphics for Alto. We'll turn on memory, and then what I'll do is paste this prompt. So create a bold kinetic typography promo video for another ad for Alto cold brew, but this time I'm saying on screen we want text zero crash. Now let's go back one more time and add our different reference assets that we have here. We'll just drag all of those in, and then we don't need to do anything else. Let's just let it run. Now we can see it's loaded the h3 visual design skill, so it's pulling from that to do the motion graphics and text overlay specifically. Now something else I want to double click on while this is loading is you can see this little browser tab right here. So right now we are in the design portal of the Minimax design platform, but if we click on browser right here, then we can see the google interface that you're probably used to. So we can go actually to websites, I can go to youtube.com, and so here we can see our youtube in here. The reason this is cool is you can do kind of research, or at least this is what I like to do it for, and pull references while I'm creating on my canvas. So let's say we're creating motion graphics, I can jump into the browser, and I can say coffee, brand, video, ads. I can go to images, and then I can start to pull from what I like, right? Maybe I like this little styled image right here, maybe I like the look of this, and then what you can do, and this is genuinely awesome feature, I've never seen this anywhere else, is on any image here or any video, I can click on this, and I can click send to chat, send to canvas, and that will be integrated then into the AI. So it's this very kind of AI native research and creative process while you're working on the canvas. Okay we can see that the motion graphics video is done, let's go ahead and give this a watch. I mean this speaks for itself, this actually is one of the most impressive motion graphics videos I've seen, so this demo happened to work out really well, and I've spent a lot of time in here already, but really you can see exactly how good the AI is with the text, right? I mean we have the perspective that's played around here, and all the text we zoom into is perfect, and looks good, and this right here is a functional complete UGC 11 second motion graphic ad. And so we're going to wrap demo two on that note, and get over to number three. And here again is the prompt that was used for that, it was literally just create a bold kinetic typography promo video, and then I gave it the text to use as well, and we had these six reference images in our canvas. Now what I always recommend as well, once you have something that you like, like let's say I really like this kind of ad, then that can become the base in the reference that you can pull from. So I can use the at symbol, I can go to video, and I can say create another five variations of this, right? Once you have something you like, that becomes the new layer of context for the AI to continue iterating, and creating more volume, and swings of that for you. How this works behind the scenes, is your minimax design agent, when you give it a prompt, goes and speaks to the video model, and can take your prompt, and make it much longer, right? Do the prompt engineering work, add the style tags, do all of that work that you used to have to do in the prompt. So it's this like add-on agentic layer, that will go and improve your prompt, to get what you want out of the actual video model, that does the legwork to execute. So there you go, that is the bulk of what I wanted to get across to you. You can get your own idea of the quality, and the output, and how that fits into your business, and what you're trying to do with these tools. And then I want to touch on limitations, and where these models, and these design agents are going. Number one, is just you continually see process in physics simulation, and being able to capture very you know, nuanced realities of physics, and how things move, and look. That's just a mainstay, these things will keep getting better over time. Number two, is length. So currently you can generate up to 15 seconds, with the process we looked at. You know, 20, 30, a minute, multiple minutes, and you know, who knows where after that point. And then three, I think part of the kind of designer intelligence layer, can distinguish a little bit better with a reference video, on what's to actually reference, and what not to reference, right. You don't necessarily want to literally copy everything about the original, you want to merge what's really good about it, with something that's maybe unique to your brand or style. That's kind of the edges of the map, and where the whole space is going. And then that 3D director stage is worth trying, if you are more of a production kind of video person, And so there you go, that's the video. Now again, if you want to support the channel, and you haven't already, you can go try this tool, by commenting minimax in the comments, or going to the link on the screen now, or in the description. Either way, I hope you learned something, I hope you can give these tools a try for yourself, and benefit from this incredible renaissance, that we're seeing in the world of AI, or creatives, and visual media. That I appreciate you all being here, and I'll see you in the next one.