How to Get Finished Video Ads for Your Business From One Prompt on MiniMax Design

So the ads that you're seeing on screen right now were not filmed, edited or scored by a person. These were generated from one sentence typed into a new AI design canvas for a new product launch of a business we've been working on on this channel. There's no prompt engineering, no separate tools involved here. This is a new way to create visual assets for your business. So ads, videos, hero images, social media images. And in this video I'm going to walk you through exactly what that new modality is and then we're going to go create all these assets for my brand so you can walk away doing the same for yours. And specifically we want to get three things by the end of this video. Number one a UGC ad with a person talking about our product. Number two a motion graphics promo so something more text and graphic design oriented. And then number three get a bunch of images that we can use for our website social media. That's going to be the video and thanks to MiniMax for partnering with me to make this one happen and more on that later. Let's get into it. Now the traditional AI workflow problem for generating visual assets is it's just slow. It's not very efficient because you're sitting there you're saying I want one image. I want one video. It gives you that say video back. There's an error at the 13th second, so you have to sit there and regenerate it for 10 minutes. And then to actually just go through this design process that you would with a human designer takes hours or days with the AI because you're just sitting there doing one tiny iteration at a time. And this new system for design solves that. We're starting to see it bubble up in a bunch of places, but I think the MiniMax platform has really nailed that with the canvas. And so that's what we're going to be looking at the solution to that traditional kind of one-on-one generation. Now you've seen me cover the MiniMax models for a long time. We've seen them evolve and get a lot better truly over time. And so the platform that we're looking at today is going to be powered under the hood by the MiniMax H3 model. It's especially good at text. You're going to see some examples of that in our motion graphics videos. And what's cool about the structure here is there's kind of H3 native prompting. So there's an AI design agent as part of the canvas, which you talk to and then it is what goes and gives the instructions to the models to actually do the work and then produce these finished outputs. And you don't have to use MiniMax models on the platform, all the frontier models are going to be available through it. And if you're really deep into this stuff, it does integrate with ComfyUI and you can download the weights locally. It's a fully open source thing, but that's not a rabbit hole we're going to get too far into in this one. So that said, we're going to jump into my computer now you will learn better if you If you follow along step by step, to do that and create a free account. You can support the channel by going to MiniMax.io or commenting MiniMax under this video to get that link. If you want to follow along, great. If not, no worries. Either way, let's jump into demo number one. So here we are inside the MiniMax design application on Mac. You're going to notice the canvas layout throughout this a lot. It's a super nice kind of node-based architecture. You feel like you can express yourself creatively. I mean, while you're doing this, I personally love the interface. We can see other videos that have been made in the gallery here, but I just want to jump straight into it now. Recently, in a different video, we went ahead and created this new flavor of our cold brew brand. That's Alto Peppermint Mocha launching for the winter season. So what we're going to do now is develop all the visual assets, the ads, etc. So specifically video for the new product launch. So I'm going to go ahead and say, create a 15 second UGC product review for the Alto Peppermint Mocha cold brew. And I'm describing what I want. And. Now under models here, we're going to go ahead and use MiniMax H3 for that model. And then we have an AI agent. So the three that's going to direct the creation of the ad. Now what I'm going to do as well as upload some reference images of the product, you can do the same thing for your product where I have some images of what it looks like and a logo file right here. So I'm just going I'm going to select these and then drag and drop them in as reference so the agents will keep the this logo and this product consistent throughout all the assets that it makes and then under select project let's go ahead and create a project called Alto Peppermint Mocha and now all our assets will be organized as well and context and memory will be shared which is very important So let's let it run now. Here's where you can start to see the canvas layout really come into action. So we can see all these different nodes that we can create director stage for building scenes and shot control tables, text, our images, obviously our video. Now the agent is starting out by asking me some clarifying questions where it's saying, Hey, I noticed a discrepancy in the instructions you gave me. Do you want to use the bottle? Or I can, I'm gonna say use the bottle as recommended. And then do we want chocolate or peppermint or more holiday delivery? Let's theme it for general holiday. We'll click submit. And now the agent will continue working. While it's doing so, I wanna touch on some of the organization that I think personally Canvas enables and I really enjoy. Let's go ahead and start by adding some text. And I'll call this notes for launch. so this can just be notes that i want to remember while i'm going and having the conversation with the agent so i'm going to leave this here now if i press on this also i write content it's basically a doc so i can say remember to vertical jamie said whatever it is that whatever feedback jamie said and then maybe i want a director stage so i can go ahead and put that right down here and if i click open you can see more of what that director stage looks like for the shot control and then now I've created a group of the reference assets here as well. So these are the assets that the agent is going to use to base the video it's making off of. Now that's organized and then there's also a bunch of edit features. So we've organized our screen a little bit. We'll put some of the videos up here and then if I press on any image or anything that's created you can see all these different tools that are available here. So Kind of Photoshop, right? That kind of modality, the canvas that we're all used to, but now really fundamentally AI native. What packaging do you want me to use? I'm gonna say, just go ahead and design a concept label. Let's just see what the agent comes up with. And so now it's loading image. And so now it's loading an image skill to generate the package and then move on to the video. Okay, it looks like it generated that image for us. Let's see what that looks like. This is cold brew, peppermint mocha, same font, colors look good. I'm just going to go ahead and say, use the concept. Now it's created a sample creator, the person for the UGC ad as well. So just someone for the home. Looks good to me as well. Super realistic and it's kind of, you know, stay at home mom vibe is I think what the agent was going for from here earlier. And now finally it's generated a storyboard for us. I'm going to go ahead and approve that and generate the ad. Now, if you want the agent to do all of that for you, you don't want to approve it. You can set that as well. If you want to approve everything you can. The whole point with these AI-native design systems is that you can be involved as much or as little as you want. If you wanna generate visuals, videos, images with no approvals, just let the AI run wild, you can do that. If you want control, You can jump like we looked at, for example, into the director stage and have this extremely granular control. So let's just wait for that final generation. Okay, the video is done. It is looking good. Let's go ahead and give this a full watch. Okay, holiday coffee just arrived. Alto peppermint mocha, rich chocolate, a cool peppermint finish, and clean caffeine. This is my kind of holiday coffee. It's a limited holiday flavor. It drops December 1st. Okay. So very good quality. I mean, you can see the text right there. Product looks consistent. Audio is good. All native. I'm extremely happy with this. This is something I could go ahead and run. And so there you go. That gives you just an idea off the bat of the quality of the generations that you can get here. Now we use the H3 model. If you want to use the SeaDance models, you want to use whatever model it is you want. You can see them all here. You can use all of those as part of the platform. And what I wanna do now is double click on the H3 model and the text generation. Again, you can see the text generation is one of the big advantages of the H3 model specifically, at least one that I've noticed in it. So what I wanna do is do a follow-up question here where I'm gonna say, create a bold kinetic kind of typography, Promo video for the cold brew and let's just say peppermint mocha. So I want big text, high contrast, 10 seconds vertical for this. I'm going to specifically again, select the H3 model. So I've selected that. And now let's let our V3 agent go ahead and create that. All that's running as well. I want to emphasize that you can kind of bulk create assets here. So I'm going to say after this, I'm going to prep this prompt and say, now create 10 mock-up images for ads. Ads for the peppermint mocha flavor, generate all of those at once. And so after this finishes, we'll let that run and you can see, you know, 10 images, however many you want, just generated at once by the agent. And I think that's really useful for the early concept kind of viz stage where if you want to just visualize 10 or 15 different concepts, pick your favorite and then explore that concept. You can do that really well with an approach like this. And I could add, for example, different concepts for each image. Now while this is generating, I want to show you one quick thing that I think is fun. You can click on browser here. This is a literal browser that is full. Okay, the video is done. Let's go ahead and give this a watch. So there you go. I particularly love the music in that one. That's actually really good. And now if we wanted a product shot in there as well, we could just go ahead and ask the agent to do that. So you can see when you click on a video, it opens the prompt. So right here I could say something like, you know, include Product shot, but overall that gives you an idea as well of the motion graphics and the text capabilities. So a lot of potential there, if that's the focus of your ads as well, or just integrating these two styles, right? So our UGC style on the left and our motion graphics text style on the right. Now we can also see those images started generate while this video was going over here. So we can see V1, this looks good. And now we can see some of those other different concept images are starting to come in as well. Here's number two. This one is fun. I like the look of that. And now if we zoom out, we can see nine images that have been generated all completely different concepts and our 10th is being worked on. So if you want to create right now is possibly the best time to do that in history for exactly the reason you can see on screen here. Okay, now let's jump into the third and final demo here, which I want to be just a deep dive into the director stage tool so we can see what we set up already. Let's go ahead and click open, then I'll click add path and I'm going to do a dolly in shot. So let's Place the camera right here. So now we can see we're gonna dolly in with the shot and the camera right there. Then I've clicked on character and prop. I have a male here. Let's go ahead and put the male in front of the camera. And I can put, for example, let's say a tree in the background, right? Whatever it is you want, you can go ahead and build that set. You can then animate the character and then we'll add a... So now here, for example, you can see as we pan through here, we have the dolly shot. in we have a human we have our tree so then if that's the information we want to convey you can make this as complex as you want then the agent is aware of what's going on so you can say something like generate a video using the setup from the director stage or you can go ahead and directly export the camera video the image And then you can feed that in as context to the so my intention at this point, right, we've covered number one, just the video quality, which I think is the highlight in terms of generation of this system, we've covered the ability to bulk generate the canvas modality, and then the nature of control that you have over generations. So how much control you can have or how little control if you want to regulate creations to the agent. So now we can look back at our handiwork here we have 20-something visuals that are created, and I think the Alto product is ready to launch. We're still a month and a half early, but the work is done, and that's thanks to Minimax Design. That's everything I wanted to cover. So I think as far as the drawbacks here are where this platform is going to continue to improve, number one is always going to be physics. That's the case with all the AI, especially video models right now. Physics are just hard, but as the models get better, the platforms will get better at integrating the physics into these final deliverables. Number two is length. So right now it's about 15 seconds max. You can stitch things that are 15 seconds together, but that's also a deal where as the models get better and extend generation limits, then you're going to be able to push past 15 seconds for your generations on the platform. Now, number three, I just noticed this when working with reference images on the platform. Oftentimes it will interpret that very literally. That's a very good thing, but if you want a, say, reference for the vibe or reference for the color and not a reference for everything, then that kind of of creative direction is something you want to very specifically give the platform along with those reference images or video or audio and then for for anything cinematic or story driven is not going to invent an insane story for you if you want something. Out of this world unique you want to just give it the story. It can execute really well. Otherwise, you're going to get kind of traditional say script for an ad. So overall across the board if you have never touched the system like this, you absolutely should it will blow your mind. You'll have a great time for anyone who wants to create lots of content and automate the production of that and this will be a fit as well and then for anyone who wants to just experiment right? So you don't want to necessarily do all the work in AI, but But you want to use it as a career Creative kind of concept visualization tool, then that is also I think who this is a perfect fit for. And so that in totality is what I wanted to show you about Minimax. Now, again, if you wanna try it out, if you fit into one of those groups, then you can visit minimax.io or head to the link in the description or just comment Minimax under this video. You support the channel when you do that. So I appreciate your time and being here watching the video. I hope it's useful for you and I will see you in the next one. Thank you.