We Tested Seedance 2.5 (Honest Review)

What is up beautiful people? Welcome back to the show. Today is the day. Look at what we've got. Seedance 2.5 has just been released. Exclusive access through CapCut. We tested it and we're going to show you the results. We've got 30 seconds native generation, advanced editing, but is any of it actually good? Because people are talking about it, but are they telling you the truth? We're going to show you our experiments today. Okay, Flux 3.0 Preview, another video generation model, is also released only through the Hermes agent. We also tested that and are going to show you the results. But it's not just that. You guys know we've been talking about Unreal Engine and MCP and its crazy powers. But so far we've only shown you videos on X. We should have tried it ourselves. Guess what we did? We tried it ourselves. You're going to see our experiments of Unreal Engine and MCP connected through Claude. And baby, it is good. Fathub, but this is not the only thing. We've got one more thing. Today is a special day. Why? We've been working on the Ultimate AI program for the past six months and it's finally being released today. Finally. And we've got a big surprise for you. So stay tuned. Shall we begin? Let's get started with Seedance 2.5. Okay, so Seedance 2.5. This is the most capable video generation model by ByteDance. We all know everyone's been waiting to try it out because of its many new features. Today, it was released, but only through Dreamina, which is under CapCut, which is the video editing app. What we did was we went ahead to CapCut, got ourselves an Ultra account so that we can get access to Seedance 2.5 and then put it through its paces. And we wanted to really see if what people are hyping this up for, is it actually legit? Is it true? Pratt, let's start showing the video. This is a screen recording that we've done earlier today. This is from the CapCut platform, right? Exactly. Okay. And before Pratt starts the video, if you just quickly go back to the beginning, what we're looking at here are three different versions of our tests. What we wanted to do, what we've recorded actually is right here. It's me in the video and we wanted to test its editing capabilities. Because not keeping some of the features and edit some of the stuff from the scene. Exactly. So when you feed it your own reference video, whether it's your real life footage or even a footage you've generated and you want to edit that footage without altering everything in it, can it actually get it right? And we tested it in different levels of difficulty. So the very first one we're going to see on the left side as Pratt plays the video now, is the video we recorded ourself here on the left side. That's the reference video. And in the video, you're going to see me grab a Pepsi can and drink it. And at the bottom there, we're going to see that in a moment, we wanted to change the Pepsi can to a Flow can. A Flow can is an imaginary can that we made with ChatGPT Image 2.0. It doesn't exist. Farah, you saw the video, right? Yep. That's a real video. So this is the Flow can that we built. Yes. And we built this for the Ultimate AI program. It's actually in there. We brought that in here to see if we can change the Pepsi can to the Flow can. Let's take a look. And you just threw the can in the air to see if the model can actually do the edit. Exactly. Farah, what is the first thing you realize? I don't know if you realize. It's bigger? No. It's bigger than your Pepsi? Look at the camera angle. Oh, it's different. Yeah. It changed the camera angle, but I still look the same. Yeah. My actions are still the same. This was really good. This was really good. And the good thing is the bookshelf behind you, everything was intact. After I was paying attention to the books, to the YouTube plaque, it was actually really good. The only thing I would say is that although the front label of the can is still very accurate, the moment I pick it up, you'll see that it switches to the side panel so Pratt can go, yes, it actually has the side panels done right, according to a reference. But as I start moving it around, the text starts morphing. Now, that is probably one of the only problems in this footage. But here's the thing. This is only 720p. Okay. I think this is the case because it's just launch day right now. And when you get it through other platforms such as Hixfield or NVIDIA later down the line through API, maybe you'll get access to 1080p and 4K immediately. And I believe by having 4K generations, we can fix some of that text morphing. And honestly speaking, I'm blown away by the fact that I look the same. The face features are kept and preserved really, really good. I actually zoomed into the video. The books. The books. I told you. I have the Leonardo da Vinci book right here. His face is also the same. And it's a different camera angle that is created. That's also crazy. That's also crazy. Did you make any other test? Yes. That is going to push the model a little bit more? I wanted to. Yes, absolutely. So guys, what we wanted to do next was we wanted to change the can, but also my body. As far as dream body. Yeah. So these are, if you have any experience with creating videos using AI, then you understand that you have to provide more and more references. And again, this is very preliminary style of testing because one of the coolest features about C-Dance 2.5 is that you can submit up to 30 to 50 references. Could you do it in Cap Cards? No, right? You can add more. You can add more. But here we just wanted to test the basics and see how well it does the basics. What did we do here? This is a different video. Same environment. But I just acted like I'm growing muscles after I drank the Pepsi can. Like Hulk gets angry. Yes. That scenario. Exactly. This is the original video. Let's take a look. Great acting. Great acting. The Academia Bart goes to... Okay. Now, before you play the next one, can you go back a little bit when we see the references? Just a little bit. So you gave a frame of your raw footage and you created this photo. Exactly. So I went to the intermediate level from beginner level to intermediate level. Went to ChatGPT. I took a screenshot of the original frame and I told ChatGPT, hey, put this flow can instead of the Pepsi can, replace them, and make sure it's a front label. And also me, give me a muscular body while preserving my facial features. Definitely doing steroids. A hundred percent. A hundred percent. A hundred percent. No, no, it's natural. What are you talking about? The chicken and cheese. Just chicken and I stopped eating gluten. That's how I did it. That's what all the freaking actors in Hollywood say. Every time there's an interview after they got jacked, they ask them, this is the Matt Damon. No, they ask Hugh Jackman. Oh yeah, Matt Damon for Odyssey. Matt Damon. I stopped gluten. The six pack came from not eating gluten. Imagine Matt Damon at what age now? 40 something. Or 50 probably. He looks great, by the way. And in the movie, The Odyssey, he was jacked. 55. 55. They asked him in an interview how he got in shape. And he was like, I stopped eating gluten. That's the craziest thing. Imagine by stopping gluten, you can get to that body. Honestly, thanks to social media. Again, nobody's going to buy that anymore. I remember when I was in high school, people would watch Hugh Jackman and all talk about how they got Wolverine body with just chicken and rice. And everybody would believe them because there was nobody to talk about this stuff. It's horrible. Now we get to see the truth. But definitely, yeah, Farazone steroids here. So once we got that frame from ChatGPT, here we actually fed that frame and said, this is what I want to look like when I already drink the Flow can. So let's take a look at what the results are. This is going to be slightly better than the last one. There you go. So same. The fact that, you know, the picking up the can and all is really, really natural. It's insane. Really, really natural. To introduce you guys to my new can, Flow. Let's see what happens. My voice is the same. It kept everything. Picking up the can. Drink it. Put it right back. Muscles grow. Okay. Your biceps are much bigger than the rest of your body parts. Look at the reference. Yeah. You haven't been training problems. I'm doing biceps and trisome every single day. But if you want to see the best version of my body, you have to wait. Because I have more. I have more for you guys. Here's the thing. I'm going to show you guys exactly how we actually prompted these. So don't worry. That comes at the end of this clip. But just looking at the prompt adherence, it got most of it right. The only thing it didn't get right is that the top and bottom of the can are supposed to be black. It's the case in the reference. You can see it. And I even mentioned it in the prompt. But it still failed to achieve that. So there are things that it still gets wrong. I still think the way I got bigger and the moment I shouted and screamed. Wasn't in sync. Was not in sync. So I was like, ah! It could be better. Yes. So let's push it further. It was like just somebody's pumping you from bottom. Exactly. It didn't look as good as it should. No. Let's go past this. Let's go to advanced. Advanced level. So when you start working on actually creating proper, I don't want to call it films, but proper ad creatives or video creatives, marketing material. Anything that can be used as a marketing material. You want to put on social media. You want to put on your website. Yes. You got to do it the right way. And luckily, because of our experience coming in from the 3D world, we know what it takes to build a scene. And if you try to bring those same fundamentals here, it helps you achieve a much better result. This is common knowledge amongst the people who are creatives, but for a lot of people who are just starting out, they don't know this. So what I did in that intermediate level is just a quick way of trying to get myself to look muscular. One-shotting everything. You try to one-shot it. Exactly. You won't even one-shot, but yeah. But here, what I've done, you can pay attention, is I went and I gave a frame of myself and I said, I want to create a character sheet and have the full body length portrait of myself and also a close-up all in one image in the clothes that I'm actually wearing in the shot. That's actually what I was wearing. And so what you can see here is exactly that character sheet. The can reference remains the same. And here, we're trying out a few different prompts. And I'll explain what each of them are as we start watching the video. So Farad, let's begin. Character sheet. Okay. So let's take a look at the first one. Hey, everyone. I wanted to introduce you guys. The first one is the exact same prompt, but now with the character sheet attached. Look at the way the muscles grow. Much better. Oh, even the scream was much better. Even your body. Look at your legs. Yes, yes. Holy shit. Do you see that? You see, the moment you understand. So again, the model is available to everybody, right? Some will use it like your first try. Some will use it like your second try. Some will use it like your third one. Somebody will even go further than this, right? I would go usually further than this. Because we were preparing for the show. I think the model capability is one thing. The way you interact and the way you prepare is another thing. And you can get very different results. Exactly. And I think that's how AI Slop is created. Because most people will try the first version and one or two attempts. But as you progress forward and as you learn how to use the tools the right way, you can actually create material that looks absolutely magnificent. Especially with tools like Sedan 2.5. Let's go to the next one. Before Pratt starts, this is starting to blow my mind at this point. I told Sedan 2.5, listen, I want you to go crazy with the camera. Don't lock it. Don't just change the camera angle. I want you to be dynamic. I want you to move around the scene. Because I want to see if my face begins morphing or not. And I want to see if the can is going to stay the same according to the reference. Let's take a look. Look at that. Wow, the can is perfect. Even the back. Look at that. Look at that. It's like a handheld camera. Look at this. Look at this. This is amazing. This was the best. So far, this was the best. Holy shit. Did you see that? Pratt, did you see that? Yeah. Look at the back of the can. What the? Yeah. And guys, the reason we all know how good this is because we have spent two, three months just designing that can. So we know exactly how it should look. It's just, it's not a random can that it can morph it or it can change it. Everything is perfect. Two details. So just want to add something here. Again, we never want to lie to you about the truth of how capable or incapable these models are. And you're going to see that throughout this episode that we're trying to actually show you guys what's right and what's not. I'm stressing the fact that everything in this video blew my mind except two things. So one, the top and bottom of the can are supposed to be black, pure black. And I know there's a way to fix it. We haven't done that, which is in the can reference, instead of just the side angle, get a top and bottom angle. But on top of that, even though I put in the prompt, it didn't take that in, which disappointed me. But the moment I turn the can, the ingredient label, the placement of it, everything is exactly like the reference. The text slightly morphs, slightly. But I genuinely believe the moment we go to 2K or 4K resolution. It would be better. We're going to have a much higher chance of achieving almost near perfect text. Next. Here, in the next video, I wanted to change it up even more. I wanted to have a completely different action. I wanted Faros to run into the scene, scream, I need flow, because that's the brand name, and drink it and tear, not tear his clothes, but how it outgrows the air. It grieves inside so much that he starts tearing up his clothes. Okay. Let's take a look. There's a surprise ending to this also. I need flow. The bottom of the can, look, is black. Wow. Whoops. Guys, by the way, we don't need AI for this. We have seen Faros like this in the studio. It has happened many times that he runs and he starts breaking things, but that was so good. What was the raw footage? Did you submit a video? No, the raw footage is still the same. Just one photo? It's still the same. It's this? It's this. This is the raw footage. The video, right? Yeah. That's the video. Wow. That's the reference video. Bro, even you standing up, it got you really, really well. Wow. That was the reference video. I didn't change the reference video. I just changed my prompt here. Wow. What I explained to it. But you see, what has happened, actually, pause right there. I'm not joking. I'm really handsome. No. But some days, when I sleep for very long hours and my hair is all over the place and I get up, I look exactly like that. Now, that's not really in the reference. It's gotten all of those details right. And when I become the Hulk, if you look at my face, I still am me. Yeah. I'm still me. It's really good. It's really good. And I've seen you this angry. Yeah. It's accurate. Pratt, it's accurate, right? Yeah, it's accurate, bro. I kept the ending as surprised because I wanted to see your reaction. Thank you. But what I want to say is I did actually put in the ending myself with text. I did explain that I want him to knock the table and shatter it to pieces as the Hulk does. And I want him to end it by saying, whoops. So the ending, when he says, whoops, all of that is planned out. But the way he carried it out is by C-Dance 2.5. Next challenge is, okay, same reference video still, but I just want him to do the same action in a completely different environment. This is a North American house, a house in Vancouver. Flow, I need flow! You're already big. No, bud. Getting bigger. Oh my God. It looks like you. It looks like you. It does look like me. Yeah, yeah, yeah. If I look at this, just go back a little bit, tiny bit, right before I knock the table and bang it. Look at the, wait, wait, right there. Look at his shoulders, his delts, and look at the soft body simulation on the delts when he bangs it. Wow. Do you see it starts jiggling? No, I think, guys, 2.5 is really, really ahead of the other models. Like, cling and all can never... No, no, no way. Let me, okay. I'll be clear about something else. Because there was a lot of talk about Flux 3, the new model. I'll show you the results of that. I'll show you the results. I want to be honest about something. What you're saying here is not what I'm saying is a great advertisement. I'm not saying, oh, this is the perfect commercial. You should put this up. Make $100,000. You're going to get views. That's not what I'm saying. I'm not saying this is that at all. And I generally still think this is AI slob. If you just put this up without any explanation, without any reasoning, unless you're a super famous influencer and people will just love the video because of your face, I don't think this content is going to work. And that's not what we're trying to promote here. What we're trying to showcase in just one hour of experimentation with little amount of prompting or referencing done, what is actually possible? Can you actually achieve great prom adherence? So the model is actually listening to you, which is yes in this case. Can you achieve realism? Realism, yeah. Can you achieve consistency? Do the physics look real? Does my hair flip about the right way as it does in real life? And I think the answer to all of that is pretty much yes. And I think, again, I always think about short commercials, short ads, marketing videos when you're making something for 30, 40 seconds. The moment you don't have those morphings of face, like having six fingers or things are changing in the background, it actually becomes really, really usable. Again, like you just said, you cannot use this for anything now, but you could place this in your pipeline or your workflow when you are creating something. Absolutely. And this is just the beginning. How much did you spend for creating all of this? I paid 57 US dollars for the ultra account. US dollars for the ultra account. And these are perhaps 70% of my generations. I've hit 30% because I was testing other stuff. How much credit do you have left? More than half? I used half of it. Half. Which is actually really good. Not too bad, actually. But, but, but, but again, again, bad decisions is always going to be honest with you. You're watching this in the middle of the screen and it's not even full screen yet for you. Once you make this full screen at 720p, the quality is not that high. Okay, right? Not in CapCut yet. Yeah, but I saw, I saw, I saw, like, I think Magnificat, Higgs, Vita, NVIDIA and all the Yeah, but the cost is going to be higher. Okay, okay. Okay, the cost is going to be higher. Nonetheless, I'm amazed at the results. Because now imagine if you actually put in the time, come up with the perfect character sheet, come up with the right references. And if you can input 15, 20, 30, 40, 50 references, imagine how much more granular you can get with the details. I'm pretty impressed. What we're going to do is not only show you guys more experiments right here on the show. Of course, we have to bring up the concept of the ultimate AI program because we're launching that today and we're going to share more. By the time they're watching this, it's already launched. Yes, it's already launched. It's already out. We are going to do that at the same time. Pratt, can we go to the website so everybody can see? Okay, so by the time this video is out, if you go to the description or if you go to ai.badestrain.studio, the pre-sale for our ultimate AI education is going to be live. Can I just give everybody a quick intro? Sure. In case you're new here, I just want to tell you what this is all about. We've been in education for the longest time. We've taught people how to use tools such as Unreal Engine and we've done it for many years now. And it's always been super exciting to us to experiment and learn all of these tools, work with clients across the world and take all of our learnings and distill it down and make it easily digestible for you guys. It's what we've been doing since day one of bad decisions. AI has been part of our workflow since I would say two years ago. But two years ago, it started out very small. We even have it in our YouTube channel when we started testing ChatGPT. Was it ChatGPT 3 or 2? Very early on. Very early on. But at that time, the reason why we never really taught you guys anything about it is because it was not yet practical. It was not yet used by us in our client work. But over the past two years, that has changed completely. We've been using it inside the workflows that we already have. And we've had specific clients where we just do AI services for them. So nothing that we knew from the past, but just AI services. And so when we saw this shift, we decided it's time for us to come ahead and create something that is just like all the other educational programs we've done in the past. And the reason why it's taken six freaking months to launch this is because the way we create our educational program is unlike anybody else. Guys, we rewrite and rewrite and rewrite the script over eight times. It's written by us. It's verified by every member and commented on and then rewritten 10 more times. Once the script is refined, then when we go into the material creation, the material creation as in recording all of the visuals itself takes another few months. At the same time, Farhad... The editing and recording and all of those. Because we had specific team members working on motion graphics. When I was talking about visuals, I'm talking about the actual content, like what we're actually teaching. We have to make it and teach it to you guys. We can't just theoretically script them. So editing, as Farhad mentioned, creating all of the motion graphics so that every 10 seconds or 20 seconds, there is visual aid for you guys so that it's exciting. Put all of that aside, we decided to create our very own platform. We're not running this education on anybody else's platform. Up until now, we were. And most other people are also using third-party platforms. I don't know anybody that doesn't use a third-party platform. We probably are the only one. We decided to create our own platform for you guys. And I'll explain why. We wanted to have the flexibility to make changes to the platform according to your needs. For instance, one of the biggest requests has been, hey, I don't have a Discord account, right? But I still want to communicate with all of the members. We created a mini social media inside our educational platform. This is it. Which is here, what you're looking at right now. So you guys can comment about your work, share your work, talk to your peers, get feedback from us. All of that is happening. And by the way, we decided to develop this platform while fucking working on the actual content, which is why we haven't slept for the past few weeks. Farah, do you want to give the guys a little overview of... A hundred percent. So the reason we wanted to really focus on this program is because AI is really different from anything else that we have taught before. AI is fast. We had to be updated and we had to make sure that whatever we are teaching you guys, it's not getting also outdated because tools are coming out every single week. And from the thousand tools that comes out, maybe only 10 or 12 is good. And you need to know which one you need to use for which task. So we created one specific section called foundation. This is your timeless knowledge that doesn't apply to any tool. You will learn this and it will stay with you forever. It is like the foundation knowledge, the theory that you should know. Then we will have playbooks. These are real life projects. These are the projects that we actually use ourselves in our day-to-day in the studio, or we get paid to do it for a client. For example, the whole can that Farah's used in the Steed Dance example is built from scratch, from research, from design DNA, from commercials, from photos and video. We created everything for that. And of course, that's not the only thing you're going to get. We also have live streams dedicated to the platform, which means we're going to have pre-recorded content, edited content with a lot of visuals, which is great. It's shorter and sweeter. But at the same time, we know that we and you guys enjoy that interactive kind of lesson, which is why that real-time talking and conversation between us and you guys while we learn Q&A feedback is important, which is why we also are throwing in live streams. We also have another module we're launching with, which is deep dive into LLMs. This is understanding everything about the strongest LLMs like Cloud and ChatGPT, but not just the basics, but also how do you actually use it? Everyone uses it for emails and asking random questions about football and medical questions, but are you actually using it for your business to help you make money? So we teach you how to, for instance, hook up your Cloud to your Google Calendar to your Gmail, have it find clients for you that will match your business needs and email them for you. How does it know which clients to find? That's another chapter. We've created a master prompt that helps you create a DNA of your brand, feed that to any LLM, and it will understand your business on day one. And I think there were questions when we were doing our Instagram Live, right? People are asking, how does your Ultimate Unreal Engine program is different from the Ultimate AI program, right? We created the Ultimate Unreal Engine for creators, for filmmakers, for people who want to learn the most powerful 3D software. However, the Ultimate AI program is for freelancers, entrepreneurs, business owners who actually want to use AI. It is not only image generation or video generation. It's not purely visual. It's actually not. It will have coding, like agentic AI. It will have, in the future, it will teach you how to build your website. Because as someone who runs a business, especially in the age of AI. Near future. Near future. Yeah. As someone who is running a business, and in the age of AI, I would say there would be a lot of small businesses. Because now we can do so many things. You need to know how to build your own application, build your own software, make your own tool, market it, create image, create videos. You should be able to do everything with all the tools that you have. And I think this program is exactly what it will do for this. Let's go back down. I just want to get to the, a little bit lower. Yes. Oh, yeah, yeah. This is another really cool feature of the program. We decided, since we have the community tab, very soon we're going to implement a voting system. Council. I'll call it council. A council, if you will. So that the members can vote on what comes next. So these are just the modules you have so far. And by the way, they're pretty long. And there's a lot of them. But nonetheless, we still wanted to show you guys what's coming very soon. So for instance, what you're looking at in those folders are exactly that. Teaching you guys how to use ChatGPT codecs, cloud code, setting up AI agents, cloning yourself, whether it's for your business or for content creation, using tools such as HeyGen. Then building websites, whether it's for your portfolio as an artist or whether you're actually providing that as a service. Or creating applications, plugins. There is a really cool video that I watched recently by Johnny Harris. I'm not sure if you guys know him. Pratt, can you quickly just look him up so people know who he is? Definitely, you have seen his video as a journalist. Like Cleo Abraham. Cleo Abraham. On YouTube. He was in Vox too, right? Not sure. Not sure about that. But he creates YouTube videos that each get millions of views. It's his own way of doing documentaries, but on YouTube. Is it Johnny Harris? It is Johnny Harris. Yeah, there you go. That's the one. Now his videos are incredible. 7.8 million subscribers. 531 videos. Each video is a documentary of him tackling a question, trying to help people understand a concept. Whether it's political, medical, technological. Take a moment to watch his videos after the show. In his videos, there's lots of visuals. And there was a video I watched yesterday where he was doing a podcast. He was talking about how he's been looking at how he can make his workflow better. Because he's been using Google Docs with like 22 shortcuts that he's custom made over the past few years. So that his 3D guys and his motion graphics artists and himself can all read the same document and still be able to use it. But he said, since I know AI has been around, I started playing around with AI tools and figured out if I can build tools. And he said, it's been so fun because I've been building all these little cool plugins all for my own team. So that we can speed up our workflow and become more productive. Just like us. How many applications we have built just in-house without selling? We have Neo who posts all of our content. We have Pigeon. Shakespeare. We have Shakespeare. So we have different, whoever wants to do something. Hooker. Yes. They will go and just build something internally. You don't have to sell it. You don't have to give it. We've done both. But we've done both. Yeah, yeah. Not only have we made plugins and applications and websites and the freaking platform you're on for the AI program ourselves. We spend a lot of time on this. Again, just like the sea dancing I told you, we didn't one shot this. This has been the work of the past three to four months since we started this program's work. We've not only used it to make our work better, we're also servicing clients with this. And we want the exact same for you. And right now is the best time. And here's why. Pratt, please go back there. Farhad, tell them about the big surprise. What we wanted to do, because we have done this for our Ultimate Unreal Engine guys and also people on this show, because we haven't actually done a lot of marketing for this program. We just talked about it on this show. We have just talked about it on this show and people who are in our Ultimate Unreal Engine program know about this. We did it previously, last round for our Ultimate Unreal Engine guys and we want to do it here again. For the period of five days and just five days, we are going to give an exclusive discount with a price that would stay with you forever. Meaning whoever comes after this five days, their price, of course, it will be increased and it might increase in the future as we add more content to the platform. It will. Because this is different. This is not like a one-time program. We're going to add more content. Yes. And as there will be more demand and more people to manage, more people to support, and as we hire more people to just help with this program, we're going to need to increase the price over time. And Farhad, I just wanted to add that. No, that's right. And for the people who come during the pre-sale period, we said we are going to lock the price. We are going to call them the founders and they're going to have a special badge on our platform and their price is going to be locked. So either they're going to pay $77 a month forever or they can pay annually, which they will get 5.5 months for free. Their annual, it goes as low as just $41 a month. That is something that we want to do just for five days. Guys, I urge you. If you've been watching this show and listening to us for however many months, maybe years, You know every time when we talk about something, whether it's Unreal Engine, Whether it was Gaussian splats, whether it's AI now, we are spending every single hour of our waking days looking at what the market's looking at, What the market needs, and how you can be better at your job, and how you can have more skills that are in demand. And you know that because you've been watching us. You've been hearing us talk about it on the show. You've been probably watching our stories on Instagram. And you know, we live both in Canada and in Dubai. So we go back and forth and we hear in different conventions what people are talking about. Whether it's at Unreal Fest and SIGGRAPH. So it's all 3D related and creative. Or whether it's at conventions that are completely business related. Since the past few months, especially, every single person we talk to, as long as they've been successful, They have some level of AI implemented in their business. Or if they don't, they're desperately looking at implementing it. They don't know how. And there's so much room for you if you want to grow. If it's for your own business, fucking amazing. It's even better because you get to use it immediately because you can solve your problems. Or vice versa, I would say for people who are just looking for an opportunity to get more clients. Because I've been there myself. We've all been there. If you want to find more clients, the best way is to get the skill that is highest in demand. And every single person who is calling us right now, at some point in the conversation, it turns to AI. It's funny. Whether it's your neighbor. Whether it's the guy at the gym. Whether it's the guy at the restaurant. It happened to us yesterday or two days ago in the lift. In the lift, yes. The guy asked us, what do you guys do? We just said we're in tech. He was like, AI? It would never happen. And I was talking to Faraz the other day about this. This wave of changes that is coming, I think it would be good for a lot of people. And a lot of people would fall behind. And I think the decision is just ours. Me and Faraz and Prat, we had our own discussion to decide, okay, which way we want to go with. Do we want to learn it every single day? Because this is different from any other tool. Like every day for this show, two times a week, we always have three or four news to talk about. This was not possible with any other technology. We are coming to you guys Tuesday and Friday. And even in the last episode, someone said, you guys missed this news. We covered three news. It was Vlog 3. By the way, whoever said that, we did cover it. Yeah, we are covering it. We are going to cover it today. We wanted to actually cover it. Did you respond to the comment? I did not. We wanted to cover it. It wasn't available for beta testers. Today, Faraz paid for Hermes agent and you got it. But going back to my conversation, it is so much happening in this space. And there are so much advancement. Every industry is getting impacted. Sometimes I could say that, okay, I'm in the creative industry. I'm in 3D industry. And this particular technology is not going to touch me. Nobody can say that with AI. If you're an accountant, if you're a lawyer, if you're a creator, if you're an engineer, if you're business owners, even worse. Like business owner, freelancer, entrepreneur, definitely is going to touch your life. And we, at one of the foundation video in the program, we talked about it. Either you are using it or your competitor is using it and is getting the job that you want. So that is why we really, really put all of our effort into it, both learning on our own side when we are delivering it to the clients and also on the education side. I'll tell you something. I'll be very honest with you. I don't think we'll ever hire anyone in bad decisions ever if they do not know how to use AI at all. If they say, I'm not going to use it or I don't know how to use it, there's a very high chance we're not going to hire them. Not because I have a beef with them or I have a problem with them at all. It's not that. It's just that every part of our workflow has some sort of AI involved in it. It's just like I say, I don't use computer. And clients come to us and ask for it all the time. So this is something that is impossible to ignore. The reason why I said I urge you is because of this. Pratt, please let's go back to the website and go to the pricing section. That we had a lot of conversation about how we want to launch this program. And I'll be honest with you, considering the amount of time we've spent, the amount of love we put into this platform, the amount of sweat and stressful nights, including the nights before where we barely slept if you've been watching Instagram. The price deserves to be way higher. And it has to be way higher considering what we want to put into it, the amount of manpower needed to run this. But just for the first five days, we said, let's just put it as low as we can. Possibly as low as we possibly can. For our audience. For you guys. Because we've been promising you guys. And you guys have been supporting us for such a long time. So we said, okay, I think five days is a long enough period. Especially at the end of the month where people usually are ready. And we've been pushing this back, I think, for a couple of weeks now. So we've been giving you guys as well time to prepare. We just wanted to make sure that you guys get in at this price. Because it's never, ever, ever, ever going to be as low as this again. This is the lowest. And the annual is almost like a joke. Because when you look at the number difference between annual and monthly, it's almost like a joke. Like that difference kind of never exists. It's like half off. It's like 48%. We just wanted to give you guys the chance to get in at that price. And then anybody else who gets to know bad decisions and watches the show later after next week will not get the same chance. And that's why I want to urge you guys. If you guys are planning to learn AI, this is the best time. If you guys have any other questions, of course, our DMs are always open. Discord, DM, email, whatever. You guys can send us messages and then we'll respond. Shall we move on? Next is Flux, right? It is. You gave it a try. Somebody commented. I did. And now. I did. And this is the concept that I wanted to run. So first video you're looking at is Hermes. Hermes is an agent. You download it. You run it on your computer. You can hook up multiple LLMs to it and multiple brains. Here it's running with my ChatGPT codex brain. And through that brain, it's now connected to Flux 3 Preview, which is a new video generation model. And I'm showing you guys this behind the scenes so that you see how it works. This is not your typical Higgs field platform right now, at least not now. The only way to get it is through Hermes. So you have to pay for a Hermes account, which is, I believe, $22. And then you get to generate videos. Now, what did we do here? We literally just gave the exact same prompt that we gave to C-Dance 2.5. Same references for the guy drinking the flow can and increasing his muscle mass. And wanted to see what the results are. Let's go to the beginning of the video, Pratt. I'll let you guys be the judge. Cut. Cut. Wrong label. Face changed. Wow. Muscle growth, but the shirt is not the same now. It's a tank top. Face change. The scream was good. It reminded me of Odyssey. What is this? Why there was so much hype around Flux? Was it the image generation or video generation? I mean, cost-wise, I think, is one-tenth of whatever C-Dance is. But again, not usable at all. At all. No, I don't even know. Like the camera cuts. I think Flux should stick to photo generation. I don't know. I mean, look at this. This part, for a moment, it was okay for like a split second. Same prompt. Exact same prompt. You didn't change anything. Look at that. The scream, though. Oh, you turned it off. Okay, no problem. No problem. No problem. Pratt has enough of your scream. I don't get why the camera was cut three times and then the can... They didn't have to. ...shifts completely. No, I think it's too early to you. Again, if you are serious and if you're working with clients, if you're working with people that they need serious work, this is still not useful. Not useful at all. It's literally AI slob. Yeah. Oh, one thing I forgot to show you guys, by the way. Pratt, can we go back to the C-Dance video? At the end of it. Oh, the prompts. Yeah. So if you're curious how I actually did this with the guys, CapCut itself has this director AI mode. It's kind of like an agent. And so what you do is just you drag and drop the videos into the canvas on the left side. And then you just talk to the agent on the right side. And the agent will help you craft the prompt. So you talk to it in plain language. And this is what I've done here. And it starts giving you the prompt, but also tells you how much credit it's going to use. And then it generates a video. And then you can just talk to it, go back and forth in plain language, which made it... Was it easy to use? Relatively, yes. Okay. Now, tell us in the comments if you guys want us to go ahead and try out all of the agents for creative AI generation. I'll be clear about what I'm saying here. There's an agent here. There is an agent in NVIDIA. Agent 2. Agent 2 that they just released. There's an agent in Hicksfield as well. Supercomputer. Yes. So these agents, the way you create with them is slightly different than the manual method, which some may still prefer. If you guys want us to do a comparison of how these agents work, which one is the best, let us know. And we'll make that happen as well. Okay. So you guys saw Flux 3.0. Okay. Opinions? No. Shit? Yeah, it's shit. I love to build. This is the reason why everybody should get the AI course. If you don't want to make AI slump like that, looking at other creators... You should join the AI course, guys. I think in terms of C-Dance 2.5 and Flux 3.0, we have a pretty decent understanding of where we're at. Of course, we need more experimentation to give you guys a better understanding. But for today, I think it's fine. We have another experimentation that we promised you guys we're going to show you. And that is Unreal Engine MCP, which is Model Context Protocol, essentially giving access to LLMs and AI brains to come into your Unreal Engine project file and be able to read and write. Which means they can work as an assistant inside your scene. Farad, any guesses of what I actually wanted to do with it? I was sitting beside you until I saw the Osaka Castle. Can you tell the guys who have not taken the Ultimate Unreal program what we actually had to do in those first few chapters? So, in order to make our cinematic, our cinematic was taking place in feudal Japan in Osaka. So, we had to make the Grand Hall of Osaka Castle. In order to do that, we had to do a lot of previews. We had to make the internal hall. We had to put all the blocks, all the columns. We had to do measurements. We had to find assets from Quixel to bring it to the scene. And I think during our program, it took us around four chapters or five chapters. I mean, if you remove the fact that we were also teaching beginners, I would say in one or two chapters. But that's about 30 to 40 minutes each chapter. So, about an hour and a half. Yes. And that was after having created it multiple times in advance to prepare for the program. Yes. This is the hall. This is the hall that we made. Yeah. But it's really dark here, so it's kind of hard to see. But you get the point. Pratt, maybe you can also help us. Yeah. Actually, yeah. This is good. This is a good view. This is a good view. Okay. This is perfect. Pratt, can you also quickly just go ahead to Google and search Shogun Osaka Castle meeting? Yeah. I remember you got your reference from there. Exactly. They are making Season 2. Actually, when Season 2 is coming out, Season 2. Okay. Shogun. This is the reference that I used with the guys when we were building the Ultimate Unreal program and when we wanted to create that meeting hall scene. So you can see there are tatami mats, those Japanese mats on the floor. There are walls covering every corner. There are doorways inside the wall. So you had to cut those through inside the program manually. And there's lots of pillars, right? And columns around. And these are placed in a specific way to create those square and rectangular shapes. And in between those pillars, on top of the wall, there's paintings. Now, all of this had to manually map out with our eyes and then try to figure out based on Google searches how big the hall should be, how tall the columns should be. And then once we had those figured out, made it multiple times, during the program, we took us about one or two chapters to sit down and tell everybody how to use the ruler inside Unreal Engine and how to measure the pillars and accurately place them, which is an enjoyable process once you get it right. But to get to that point, it's quite time consuming. And if your goal is not just to create that asset itself, it's probably one tiny part of the entire workflow. So we said, okay, Claude, technically, if it had access to Unreal Engine and has access to the internet, should be able to look at this reference image and go and research and figure out how big, width, length, and height this room should be. It should be able to figure out where the pillars are, how big the pillars and columns should be, how should they be placed, and then it should also figure out what kind of materials to put on them. That's exactly what we asked Claude. Let's watch it. What did you use? Fable 5? Fable 5, the most capable model. And what you just saw, pause, pause, pause, let's pause. Of course, I sped up the video. And of course, we discussed this. We're not going to show you guys all of the prompting in the beginning. But it was, this first part was just one prompt. This first part, until here. Which was exactly what I just told you guys. Go figure out the Osaka Castle interior hall. Place the pillars and the walls. And then, we said, okay, the pillars are placed wrongly. If you look at the reference, there were no pillars in the middle. Yeah. Right? They were around the corners. Go ahead and look at the reference again and try to rebuild it. Let's see what happens. So you rebuild it. Yeah. Let's see what happens. So every time the response took about 10 to 15 minutes. Boom. Pause. Look at that. So now you have those same square shapes on the wall. The same rectangular shapes on the wall. And it's added the material. It's crazy. Did it get the assets from Quixel? I got the assets. You got the assets. It cannot get the assets. It cannot get the assets. It told me what to get. Okay. But it said I cannot get it. There's no API for Quixel Megascans. You need to download these assets for me and then I'll do it. Okay. Wait. Question. So when you say you got the assets, is the thing that's on the screen, are those also assets or those are default cubes? So if you go back a little bit, those are the default cubes. And then... The texture is the asset. And then it added the textures. Yes. So here it's added textures. It's not the materials yet. It's not entirely correct. So now it's asked me to go to Quixel and you'll see me do that. But man, how much time this would save? Okay. So I go to Quixel. Okay. I find the materials. It told you what to take from Quixel or you knew? I mean, you knew. I knew exactly what I should take, but it told me. Okay. So look around. What do you notice? You saw the... Go back. Go back when we go closer. Yes. Go back to the wood. See, wood material has been applied to the pillars. Even the UVs are correct. What the fuck? Yeah. Mm-hmm. It's good. And look at those walls. Again, I remember when I was making this with the guys, I was like, guys, should we do Boolean into the wall? Or should we have two walls and then another wall at the top? None of those decisions had to be made by us. We just told it what we want. Yeah. You could spend more time. I remember initially when you wanted to make that cinematic, you had a lot of different ideas. I wanted to rain. I wanted the scene to be like 10 minutes longer. I could have done so much more. You could have done so much more. I could have done so much more. You could have done so much more. Now, let's see. Okay. So this part, I do apologize. When it happened, I wasn't recording. Apology accepted. The moment it actually made the change, I wasn't recording. But I want to tell you guys what exactly happened. So I told it, if you go back a little bit, Pratt, if you go back a little bit, I told it about its mistake. I said, okay, the tatami mats are down, but they're not arranged correctly. You are supposed to see that gray grid box around every single tatami mat because they plant them down, you know, I don't know, in rows or one by one. So that's wrong. And I said, the walls need to have paintings. And it said, but I don't have a material for the paintings. I said, go find it. Can you go to internet and bring photos? It got the paintings from Wikipedia. Oh, ugly then. No, no, but Wikipedia has accurate photos. But play it, Pratt? Why it looks fine. No, it stretched them the wrong way. It stretched them. Look at the other wall. I go closer to it, Pratt. You can go. It stretched them. Look at this. Stretched. Okay. Again, it's a great starting point. I think you can imagine, for example, you're building a whole city or you're building a huge scene. You ask it to build everything and then you go fix one by one. Yeah. Did you notice the floor? Go back a little bit, Pratt, please. Do you guys notice what happened to the tatami mats? Wow. The distance between the... They're arranged. Again, not perfect. Did it use motion design or no? I don't know. I don't think so. Oh, it's exactly. I don't know. No, I didn't have time to check. Not that. It even put the stage here for where the Shogun sits. That's so cool. There was a stage? Yeah. Look at it. We didn't have it. No, no, no. I see that stage. I'm saying, was there a stage or is that a mistake? I can't tell. I think it added it on its own. Did you give any references from our course? No. No. I did not. I did not. Can I also add one more thing? Yeah. Okay. Take a close look at the haze and the atmospheric fog. I asked for that and it added that. Oh, shit. Even the... There's a bit of volumetric fog in the scene. Nice. No, again, it saves so much time. And I think for someone who wants to create big environments, big scene, games, it would save them a lot of time. Dude, I just thought about it. Imagine you guys were not here and I was just like a solo artist, right? And I was delivering for a client or even working on my own content. I don't know. I had no help. Imagine I had no help. And I can imagine most people are like that. And I have the skills to use Unreal Engine and I have these big ambitions of what I want to create. Exactly what I was telling you when we were building the program. You just mentioned it. I wanted to add rain on top of the temple. I wanted to have a water dripping off the rooftop. You wanted to have the whole yard. Like a huge yard. We didn't have the time. Because I remember we spent so much time figuring this part out and making it easy to understand. Right now, if you were to redo the program from scratch, I would start out the first two chapters by telling them how this should work, showing them the basics. But I'm not going to build the whole thing. I'll just say, okay, so the MCP is here. If the course is for beginners, of course you want to teach that. But if you're just teaching people a tutorial of how to build a scene, of course the first part can be done with Claude. Everything after here, I would assume you kind of have to take over. But just this basic part of setting up your scene, god damn it. Is this incredible? It used 2% of our Fable credit usage. We're on the max plan. That's 20x. So 2% is still considered relatively... You don't need to use Fable. Probably you could do this with Opus. I think Opus would make dumb mistakes though. Yeah, but you could fix it. But you bet your ass this is going to get cheaper. True. And better LLMs are going to come around. So it got cheaper. And this is going to be something that everyone has used. Why are we also showing you guys this? Going back to our discussion earlier on. Again, we're not going to sit here and sugarcoat it. We love both Unreal Engine and 3D. And we also love AI tools. And looking at how they're marrying each other here is frankly exciting to me. It's beautiful. It's exciting to me because I know how much more we can go ahead and do with this. 100%. And I don't know about you guys. Let me know what you guys think. Let us know in the comments. I'm generally excited. We're going to use the MCP in Unreal or Seed Dance 2.5 when it comes out on Tuesday. Maybe by Tuesday Seed Dance is out on different platforms so we can do more experiments. Hopefully 4K resolution. Yes, even 4K resolution. So we have to go. Yeah. Because we have some launch to do. By the time you guys see this, it's already launched. But after the show, we're going to launch the Ultimate AI program for you guys. I just want to say thank you to everybody who is watching and listening on all platforms. Spotify, YouTube. Thank you guys so much. You guys are awesome. And thank you for joining the Ultimate AI program. We hope to see you guys inside there. We're going to go ahead and launch it right now. Until next Tuesday, have a great weekend. Ciao.