NEW Unreasonably Good AI Video Generator. (What's Sora Anyways?)

So folks, it appears that a brand new AI video generator is amongst us. I've seen little clips of this one on and off in my Discord server, which you should join by the way. And there is something different about this video generator. I'm gonna start off just by pulling some clips from my actual Discord server created by fellow users like you. And yeah, the generation quality here is pretty darn high. Of course, we're nowhere near perfection yet, but this is definitely competitive with the original Sora Dem that we saw. And those were also likely cherry picked. Here is a very different type of generation. This time we have an animated character and it's able to keep the character very consistent. You can see the hair of the character slightly moving in contrast to the face. We can also see the lighting bouncing off the eyeballs of the character. It's like this green light coming from the left hand side and it looks pretty darn good. Very sharp, very clear. And brace yourselves folks, because here's where things get really crazy. If you try this thing out with some animation. In this example, we Batman, you can see that the quality levels here are really, really astounding. I mean, take a look at this. It's almost some near perfect, dare I say, Batman animation. I mean, there's definitely some AI weirdness in there, but it's absolutely better than pretty much everything else I've seen, bar none. At least for the animation style of things. I mean, this really, really surprised me. It has the frame rate down of like comic style animation. It's really crazy. And remember guys, all of these clips are just from regular. This isn't directly from the company or anything like that. It was generated and then just sent into my Discord server. And thank you very much by the way to the users on my Discord server for sending these in, because they are pretty incredible. This clip in particular shows some vfx. You know, he's grabbing some sort of like technological thing and moving it around. I don't know what you would call it, like a floating blue computer window. And that looks okay. That's typically still pretty difficult for these AI video generators to figure out. But his face looks pretty, pretty darn good. It's very consistent, it's very realistic. It doesn't morph too much as it moves and speaks. And also you'll notice his arms, clothes and hands do a pretty decent job as well. You can even see the veins on his arms consistently stay in place as he moves his arm around. And that's like pretty mind blowing to me at Least not to mention, I mean, this shot looks pretty cinematic. I think we can agree usually the cinematic style. The different styles are pretty easy for image and video generators, but just thought I'd point it out. Looks pretty good in this clip. You can see there's a little bit more of an advanced camera movement going on. It's sort of swirling around these two characters heads. You can see the light sort of moving on their body armor or whatever you would call it, spacesuits. One guy isn't wearing a helmet, which is a little concerning considering they're in space. But thankfully the other guy is wearing a helmet. And yeah, I'd say the overall quality is pretty darn good with this as well. The movement is very, very natural for the camera swinging around and the movement of the characters subtly talking to each other. Quite decent. You can see this thing was definitely trained on some movies and as we saw with the animation earlier, definitely also trained on animation. It's got a lot of those little nuances, those little natural nuances down, and it's honestly a little frightening. This is a little bit of a more stable clip, a slow clip, just zooming in. Looks like maybe it's a cutscene from a video game or something like that. But you can see there's a lot of little details in the dust around this character, the glowing dust. And then also everything is very, very consistent with only the backgr slightly changing and more contrast coming into the image as it zooms into the face. So again, very temporally consistent, which is impressive for video generators. Now, honestly, I think this. This was the clip that really blew my mind. I saw this this morning when I got up and checked my discord and I'm like, okay, this thing is serious. We've got a pretty decent rendition of an iPhone here. And then we've got our classic YouTuber. And it looks like a pretty professional video as well. If we actually play the clip here, you can see, yeah, it's a pretty stable camera here, and nothing's jittery. He's just sort of holding this iPhone and talking about it as if he was reviewing it. And his face is really clear. The smaller details in his hair, you can see some shimmering, some of that weird AI shimmering with it, trying to keep it stable. But his face looks remarkably good and his hand movements are all too natural from what you see in, like a typical Tech Review YouTube video, which is just, again, it's pretty creepy. And he's also wearing a lapel mic. Like, that's a really interesting detail. So, yeah, not only is it trained on movies, animations, it's also clearly trained on YouTube videos. So his hands also look good. He's not having any extra fingers. He's kind of just moving, like, pointing to it. It's. It's just so natural in, like, these really weird, nuanced ways. Again, borderline on the creepy side, because it's so, so good at that. And to top off our discord clips, which were all sent this morning, by the way, we've got something a little bit more advanced. This is a car running over an Xbox. You can see there's, like, four Xbox controllers, and if I play the video, it does an okay job. We've got, like, some broken. And the car is kind of like backwards and forwards at the same time. I'm not sure exactly what's going on. It's like running over it backwards. The wheel looks okay as it turns. Again, that is something that is notably difficult for AI models, is to make a consistent wheel that wraps around like that, moving. Got the Xbox logo in there, and it definitely looks like a pretty decent Xbox. Although, again, the destruction, I think, leaves a little bit to be desired. This is a difficult prompt, though. This is a difficult test to really push the model and see what it can do in the realm of physics. But again, also clearly trained on YouTube videos, because I know there's a bunch of YouTube videos of people just running stuff over with their car. So what's the story behind this model, and where can I use it and how much does it cost? Well, we are going to dive into all of that today. That is the goal of today's video, after all. But first, a quick message from today's sponsor. LTX Studio is finally public. No more wait lists. Plus, they've added some incredible features. First up, we got face motion capture, which is just wild. You can use your webcam to record up to 30 seconds of your face doing actions, and LTX Studio will take those actions and mimic the gestures on an AI character. They've also got character dialogue, and this works in both the storyboard and shot editor. You can actually now assign specific dialogue to specific characters. This map will lead us to the sacred golden lemon. They've also got a new canvas feature, which actually allows you to use brushes to paint your scene. Use a bigger brush for more pronounced results, and try to aim for something that looks 3D. But remember, AI is going to handle most of the legwork. Keyframe control is also a new one. The more movement and motion you add, the better the results. And finally, They've also introduced team collaboration, so now multiple users can work on the same project at once. What really sets LTX Studio apart is its focus on the pre production process. It's not just about generating a nice looking video, it's about actually taking a creative concept and bringing it to life. Taking you from an initial idea to the final video. The best part is that LTX Studio is offering a free trial of one hour of compute time, which is definitely enough to get a good feel for what LTX Studio can now do. So if this is in your wheelhouse, head over to the link down in the description below. Give it a try and let me know what you think down in the comments. And hey, if you create something cool, please share it to my Discord server. Huge thanks to LTX Studio for sponsoring today's video. Now back to your regularly scheduled content. Welcome back everyone. Thank you for sticking through for the sponsored segments in my videos. It is literally the reason that I am able to pursue this channel full time. So if you click the link down in the description below, it's going to take you to this website, which is hell uai.com and surprise surprise, this is a Chinese AI video generator, just like cling AI, which is probably why the quality is so good. They're just ripping training data and just tossing it right in there. No bars held. At any rate, they have some sort of like ChatGPT thing. If you have questions, why not ask the magic conch? I guess so, but that's not what we're here for. We're going to want to log in now. You do need a phone number, but you can use it for free. And that's kind of the dope part about this. I won't show you my phone number on screen right now, but you can indeed send yourself a verification code. And I'm in the US and it works fine with my phone. So you don't need a Chinese phone number. It seems to support a lot of different countries. And boom, there we go. Login successful. Now I can click this little button which I'll also link in the description below if you're already logged in. And this right here takes us to the AI video generator, the Minimax generator. Let's try a cinematic view of Earth. You can see now the video is generating and if I switch to the Mine tab, you can see that, oh, there's 137 people ahead of you. You'll have to wait for five minutes to get the generation. So generation time right now is incredibly slow. You're also going to notice that you can browse through the featured videos, which we're going to do while we wait for our video to generate. And you'll also see a shocking lack of opt. There's no motion options, there's no imaged prompt options. So no beginning or end image. There's like nothing there. It's just a text field, which is pretty crazy considering the quality and especially the Batman footage that we saw earlier. I mean, I thought for sure that that had to be some sort of image to video, but nope, all of this is generated with text prompts. And yeah, it's all pretty incredible. This is the Minimax generator in combination with Halu AI. The movement here is incredible. I think the explosion is pretty great. The fire looks fantastic and the old warehouse also looks pretty fantastic. There is some melding going on over in the corner as the explosions happen, but that is neither here nor there. I think in comparison to the other video generators that we see these days, this is pretty impressive. I mean, look at the movements on this creature. That's something that's really hard to do right now with even Gen 3, which I would consider to be among the best of the video generators we have available to us right now. And the crazy part is this one is free. However, of course it takes forever to actually generate anything here. We've got a video of a kid playing a video game at home and then the character comes out and comes to life in his bedroom, which is pretty crazy. This looks like again, something you'd straight up see in a kid's movie. I really like this generation. The creative description here is just 19th century man packing his bags. He's simply moving out. He's got places to be, but I really like the place it put him in. We've got this old clock in the background. It looks like a scene from a TV show from the 19th century. And absolutely, he looks like he's kind of ruffling and packing his bags in this old suitcase that they would have in the 19th century. Absolutely love it. Pretty simple prompt and it did quite a lot with it. This is a pretty crazy test as well. We've got pixel art style, 8 bit canvas with a kitten walking down the busy street. But it seems to really want to do 3D for some reason. But this also has a little bit of a text test going on. So we've got September 2024 displayed on the screen. It kind of followed it. It definitely got 202024 right. But September is not perfectly spelt, so that's probably something that they want to work on there. Obviously though, this thing is not really ready for prime time. It's sort of a tech demo, if you will, because it is completely free, it has zero options, and it's not like this advanced interface that you get with gen 3 or cling, etc. And even those are still in development. Here's another pretty shocking one. We've got a camera just sort of panning up from the woods down to a little girl and she's got some sort of woodland creature and she's got a bunch mushrooms in a little bag or something like that in a little carrying case. And again, very cinematic shot. I love the camera movement slowly panning down to reveal the characters. It's exactly something you'd expect out of a TV show or an advertisement or a piece of cinema. Again, it's not all the way there yet, but in comparison to what we have right now, it's pretty darn shocking. Here's another one. We've got a dark haired mermaid swimming on the water surface. And this one is in a little bit more of a slow motion type of deal. I don't know why AI video generators constantly do this, but they always have slow motion physics or slow motion scenes being produced when not actually asking to be produced in that way. Literally every video generator does this. I don't know why it's something with the architecture or the training or something, but yeah, here it is. It's still not a bad scene. It's just in slow motion. Theoretically, if you had a longer clip, you could just speed it up and get your regular motion. Here's another pretty mesmerizing one. We've got Godzilla walking through a jungle and the way he steps and looks around is pretty awesome. I really like the consistency on this one. All of the trees and all of the plants, they don't really move. They're staying where they should and you can track them pretty effortlessly with your eyes and kind of understand what's going on. Again, on the finer details, you can tell there's a little bit of AI weirdness, a little bit of AI movement, but overall I think Godzilla looks pretty good. And he's also walking correctly with his legs and everything. They're not mushing together and mushing with the ground and mushing with the rest of his body. He just looks consistent. And that. That's a really strange thing to see in the AI video generation space. Especially considering the fact we had nothing even close to this, not even close before this year. This has all Come this year, it's been a year of huge leaps and bounds for AI video generators as a whole. And this has been kind of the cherry on top of everything so far. And we're just getting into fall time, which is crazy. So there's probably more to come. Hopefully Sora comes out this year, and I hope it's at least as good as this, if not better. Here we've got the classic eating. Here's a guy biting into a nice burger. All the food is really consistent, though. I don't really see too much change going on in the background as the video plays out. And this guy is kind of just taking a nice bite and enjoying his burger, more or less. This right here is a pretty complex shot. We've got a dragon flying out of clouds, soaring around an ancient castle. And, yeah, you can see there's a little bit to be desired in the movement here. I think the dragon's wings actually look pretty incredible, which is crazy. But the rest of the body of the dragon is. Is maybe a little bit stiff, not really moving as much. So it's got some work to do there. But that castle in the background with the clouds, man, it's pretty consistent. It's easy to track and understand what it's shaped like just from the camera movement. So these video generators, again, they're really, really getting somewhere in a short amount of time. Shockingly short. I thought that getting to this quality level would probably take until 2025, but here it is in, like, little over halfway through 2024 or something like that. AI moved fast. This is a pretty great video as well. We've got some cats eating some fish and some dogs eating meat. And then we have monsters coming through the window, apparently ready to attack these animals. I would hopefully prefer that they're here to adopt them, personally, but it does not appear to be the case. However, I do like the consistency on the animals. Again, that's not something that's always super easy to do. Usually these things are trained specifically to be good on humans, and they kind of ignore the animals a little bit more. But these animals all look pretty darn good. I actually don't see any cats melding too much with the dogs. Maybe the ones in the middle are sort of mushing into dogs. Those cats are transforming a little bit. But overall, again, really impressed. And don't even get me started on the camera movement as a whole, it's nice and consistent. Here we've got more impressive animation. This is obviously supposed to be that Japanese anime style. We've got a 3D background with a 2D character. Looks alright. Not as impressive as that Batman one I showed you in the beginning of the video, but overall still pretty good. We've got a nice video that's supposedly from the 1950s of a robot walking up to a restaurant or something like that. And I gotta say, the leg movement here, the arm movement is awesome. Again, that's something that is traditionally pretty difficult to do with current video generators. I. We saw a lot of good leg movement in Sora, but Sora is probably cherry picked. It's not released to the public. No one has access to Sora yet. We have access to this thing. It's actually free. And the leg movement is looking chef's kiss. This is really damn cool though. This is another one of those shockingly impressive ones. This is camera footage, fisheye lens from like a ring doorbell or something like that of just outside your home. And we've got a zombie walking up and looking into the camera. And this is like creepy realistic. Everything is so stable in the background. And the lower video quality footage kind of makes this feel more realistic or feel as if it could have really happened. And he kind of just pokes his head in there in a pretty scary fashion. There's also a shocking amount of detail on the zombie himself. So again, this thing is packed with detail, it's packed with realism, packed with good movement and good lighting. It's. It's a really versatile model. All of these demos have proved that. I think it's intelligent. It understands movement, it understands objects in relation to each other. And it's. It's one of the more competitive models that I've seen so far. We've got this little cat walking like a person. This is honestly downright creepy. And then we've got the dog also kind of trotting in like a super realistic way. Look at the fur and look at like the body fat of that wolf. Sort of just sort of like jiggle around as he walks. It's just very, it's very realistic. And again, it freaks me out. It freaks me out because it's like, dang, we're already at this quality level. Like we're already accomplishing these tasks because I thought, I didn't think we'd be this close, but here we are. We've got access. And here is my cinematic drone shot of Earth. I think this came out pretty decently. It's temporally consistent, like I said, packed with detail. Maybe not a shot of Earth from space like I wanted, but I should have specified that in my prompt, to be fair, the clouds are moving a little bit fast, but I think that's sort of for a cinematic effect. I did specify it to be cinematic after all. Look at this. This looks like a scene straight out of a movie. Look how good his hands look too, man. And the lighting and everything. I mean, that's just crazy. Victorian bedroom here and then it. Oh, there's the cat. That utilizes the bedroom, of course. I mean, there's. There's not much else to say other than this is possibly the best quality video generator I've ever seen, with the caveat that it's super slow and doesn't have any image to video. And also, this website definitely leaves a few things to be desired. Now I'm generating Man Bites into a Juicy Lemon. Wow. This is impressive too. So we've got like a bustling metropolis and we've got cops chasing down some bikers. The first start of the scene looks really great with the camera whizzing by with the bikers, and then the cops are kind of all over the place. But still, compared to what we have access to right now, in compared to those video generators, this is like a staggeringly good shot. It's a staggeringly good generation. Like, oh, man, this is. This is where we could start creating AI films, like, if we had access to this level of quality. And pretty quick. Like, I feel like this is borderline watchable, for sure. A few improvements too. Like, next year, you're absolutely gonna see short series made fully with AI and they'll be very watchable, I imagine. Right? Do you guys disagree? Do you agree is the prepared for stuff like this? Definitely not, right? We're gonna see some crazy stuff going on. Alien the Movie poster, but with Muppets. Okay. I actually kind of like that. This is nice. It just sort of zooms in. His legs are a little weird, but it's definitely sort of talking like a Muppet. Like, that's exactly what you would expect his mouth movements to look like. I think what I'm gonna do right now is do a bunch of my own generations and tests and then get back to you. And by get back to you, I mean you're gonna see it in like two seconds because I'm just pausing the video and then generating some stuff and then going over. See, I told you it would only take two seconds. Anyways, I've actually prepared a small little comparison for you. I think a very deep dive comparison between all of the video generators is in order at some point. But since this is Just like a first look, we're gonna do a small comparison to RunwayML, which is my favorite generator, the Gen 3 Runway ML generator. So obviously this was my first Minimax generation cinematic view of Earth. I put the same exact prompt into Runway gen 3, and I got got this video, which I honestly think is also a very cinematic shot of Earth. And comparing the two is a little bit difficult. They're very different. This one's from space, like I expected, so I would honestly be inclined to give Gen3 the win. But there is a little bit of weirdness where it's like space and then it overlaps and transitions to Earth, which could be cool, but not exactly what I wanted. I didn't specify that, and it generated it anyways. So I think it's kind of a tie between both of these. This one from Minimax is obviously a little bit more on the realistic side of things. Next up, we've got man bites into a juicy lemon. And you can see this one is quite ferocious. It's definitely creepy and frightening, but he is totally biting into that lemon. He's mowing down on it. And yeah, it's like juicing for sure from the bottom, but it looks like a jelly. I really don't like the way it looks, but I mean, he's biting into a juicy lemon. And there's no doubt about that. It followed prompt and it's a pretty realistic video overall with lots of detail on, like, the skin and like the guy's face and stuff like that. It's gross, but, I mean, it's what I asked for. Now Runway Gen 3, I think it also did a pretty great job. Definitely a juicy lemon. I think the lemon looks a little bit better, but you'll notice after the fact, gore warning. He seems to kind of bite into his own lip in its bleeding and the juices of the lemon. And like, it's. It's gross at the end. I think that the first generation with Minimax is my preferred one. Although this one's also pretty good in some ways. In some technical aspects, it's. It's a disgusting generation, though. Low key. Like, I don't like it. Both of them were gross, but this one's like gory, man. Why? Okay, and this next one is supposed to be a little bit more complicated. This is 2D animation. Wizard casts a spell using purple magic on an elephant, turning him into a cat. The camera zooms in on the cat. So we definitely got our wizard, not 2d animation. Definitely decided to just go with 3d animation outright. Not sure, why that's the case. We got the purple magic and he's casting it on the creature. And the creature is an elephant, as we expect. But the elephant never turns into the cat. So it kind of drops the ball here at the end. Maybe if it could generate more than just a couple of seconds, we could see it turn into a cat. But didn't exactly get all of the way there. And you think Runway Gen 3 might be a little bit better. But the answer is no. We've got like a wizard, kind of a wizard. I don't know. He's got two faces a little bit weird there, casts a spell with purple magic, which looks great. The 2D animation looks great, but. And then it does turn into a cat, but it kind of just turns into a zoomed in image of a cat. We never see the elephant firsthand. And yeah, it just kind of turns into a talking cat. So I don't know, I could kind of go either way. They both missed a lot from the prompt and didn't really do a good job. So next up, this is probably my favorite prompt that I, I had going today. This is YouTube video YouTuber showing off his new 4D hyper dimen Tesseract. He is casually reviewing it and yeah, definitely captured that. He's got this tesseract that's sort of just transforming and moving around. It actually says 4D hyper dimensional on the tesseract when he holds it up to the camera. But I like that he holds it up to the camera and the camera tries to change focus to focus on the tesseract. I think that's pretty cool. I think that's a very YouTuber Y thing. Like, isn't this right here something that YouTubers do quite often? They hold something up to the camera and then the camera changes focus on it to show you the detail and then it moves back.