VEO 2 is now Publicly Available! Hands on Testing! (BEST AI Video)

Folks, this is not what I was expecting to make a video about today, but I'm pleasantly surprised Google actually gave US access to VO2. I know it's a shocker. For those of you who aren't caught up, VO2 is the best AI video generator on the planet and it has been for the past two months. No one has been able to lap it. However, it's been non public for those two months. Only a select few beta testers have got access. But now finally it is public. And this is the full Google VO2 model. This is not a fine tuned model for YouTube shorts. That's lighter weight and not as good. This is the real deal. And coming to pretty much everyone's surprise, it's available on Freepik. Check out this announcement. VO2 is here first on Freepik Worldwide. I mean, I had to rub my eyes and do a double check here. Excuse me. On FreePik, Google partnered with us to debut the most advanced AI video model in the world. Unmatched realism, precision and smooth animations. What a surprise. Honestly, I'm shocked because I would think that Google would debut VO2 on its own webpage. It's pretty much what they've done with all of their other technology that they've wanted to make public. They've made it available through some Google app or website. But not this one. No, this one's going to debut on freepik for some reason now this is going through Google's API. It's not being hosted by freepik and the servers are getting clobbered right now. But hopefully we'll get some generations through today. I really want to test this thing and play around with it. I also assume that freepik is just the first of a lot of other AI video generator companies that are going to be debuting the VO2 API on their site. I can imagine sites like Korea, AI Holo, AI might get access pretty soon. Obviously though, all of these clips, I mean, they're downright insane in terms of the quality level. Google VO2 beats everything out of the market, no questions asked. Now you'll see here that the first 10,000 users get two free generations. This is already used up to the best of my knowledge. And when I tried to cash out those free generations earlier, I was met with a nice little screen here saying that VO2 is at full capacity. I'm also going to be showing you guys a free way you can use Google VO2 with a special code that I've just been given. I just got a DM with A code so you guys can get some free credits for VO2. So that'll be in this video. But before we talk about pricing and all that stuff, let's actually try to generate a Google VO2 output. In the creation tab under Generate a custom video, we can select Google VO2 as our model. We can either do 16x9 wide or 9x16 just widescreen. Flip the other way around for a phone or something like that. And 5 second max duration for generated clips. Even now that it's public, it's still quite limited in terms of what settings you're able to access. First prompt scene from a 3D animated feature film. 3D animated lemon character, Pixar style is eating a cheeseburger in a restaurant. And we'll click generate and on my plan, I have 18 total generations available to me using Google VO2. Yeah, it's pretty expensive. We'll go over pricing and credit comparison later in the video. It's not cheap though. Let's try the other aspect ratio of 16 by 9 for our next prompt. A first person POV video uploaded to the web recorded on an iPhone meeting a green alien in the middle of the woods on the east coast. We shake hands. Lovely. Send it through. Let's test it out on video game gameplay footage. Minecraft gameplay footage. Building a snow golem in Minecraft. That's something that you could probably see in five seconds. So I'm gonna send that through and you can see things are kind of just building up in a queue. And hey, our first generation is now done. Oh, there he is, our 3D Pixar animated lemon character. He looks alright. He's definitely got a burger in a restaurant, but he's sort of just talking. He's not actually picking up and eating it. Very interesting indeed. I mean, it's very good quality. It's very coherent. It's definitely 3D animation and I like how consistent the hands and the character are. It's just he's not eating the burger. Maybe if this was a 10 second generation, he'd eventually pick it up and take a bite or something like that. But as far as a generation goes that didn't follow the prompt to a T, this is pretty good quality overall. Oh, looks like our alien video has also completed. So we got the first person POV with the iPhone. Okay. I think it didn't really capture the idea that I wanted, but it makes sense given the prompt. This guy's literally just kind of fumbling with his phone trying to take a video of this alien. Yeah, First Person POV video uploaded to the web recorded on an iPhone. So it's trying to put the iPhone in the POV as well. And it does that pretty well. This is not the first video generator I've seen to take a prompt like this and include the iPhone as a part of the prompt as well. Overall though, I mean, it does look pretty clear and it does look very much realistic. We can even see the person's reflection in the iPhone. If you look very closely, you can see that it looks like there's a face or person there in the reflection. Now we have Minecraft gameplay footage. Oh my God. It's decided to go with Minecraft Pocket Edition with the hand controls and everything, which is pretty crazy. This is staggering. It's definitely like a little snow golem. It kind of looks cool, but yeah, there you can see it is trained on plenty of YouTube data because it nails the Minecraft interface the correct amount of slots. It's got the PlayStation controls down here that looks like some wheat, some snow blocks that's supposed to be a bed. But like, even the UI is pretty darn close. Pretty crazy to see something like this. You can tell it's trained on some insane data and the textures are so close. Before we dive into some more VO2 generations, I've got a quick word from today's sponsor. Huge thank you to you guys for putting up with sponsors. They quite literally make this channel a full time job for me. They make that a reality. Today's video is brought to you by InVideo AI v3. This is the only AI video tool that makes you full length ready to publish videos using nothing but a simple text prompt for a special demo here I'm going to showcase the hyper realistic films workflow. Check this out. I'm creating a three minute long generative film for YouTube about Oppenheimer's legacy with Einstein. This custom prompt is asking to tell the story of how these two legends first met and how Einstein supported Oppenheimer through his journey. We also specifically request InVideo AI to use details from a particular article which we will link and check it out, guys. I think the results speak for themselves. The pursuit of knowledge is noble, Robert, but we must always consider, consider the consequences of our discoveries.

I didn't fully understand his warning then.

How could I?

The Manhattan Project consumed me. Day and night we raced against time and our own consciences. Einstein's letter to Roosevelt had set it all in motion. Yet he remained outside the project.

And on the plus side, if there is something that we do want to change about this. They have a really easy to use and seamless editing capability. We can literally just use natural language to ask what we want to be changed in the video and the AI will go ahead and make those changes for you. If you want to be extra precise though, you can use the built in editing framework to tweak any piece or portion of your project. If you're ready to take a dive into the cutting edge of content creation, click the link down in the description below to check out InVideo AI v3 for yourself. It's quite a surreal experience. Huge thanks to Invideo for sponsoring today's video. Now back to your regularly scheduled content. Welcome back folks. Let's get back into it. Okay, I want to try to fix some of these generations. Since I know this is trained on a vast amount of YouTube data, we're gonna try this prompt POV. I met a real alien in the woods. His name is Glorp. So we're just gonna title it like a YouTube video and send that through and see how it does with a more natural or YouTube title prompt. Maybe that will, you know, hit the right chords in the training data, so to speak. Well, now that we know it will try some copyrighted stuff, especially this Minecraft gameplay. Let's see if we can get some other famous characters in here. Cine Wedding ceremony, Shrek and Bigfoot get married. Yes, that is what I want to see. Tech reviewer, YouTube video throwing my gaming computer into a volcano. That would definitely get a lot of views on YouTube. Linus tech tips. Better get on that. So for this next prompt, I'd like to enhance it close up macro video of a small mushroom village teeming with life. Hence the prompt with AI. Obviously it's using some sort of LLM here to rewrite my prompt. Give it maybe a little bit more detail and life. Sure, that's a very detailed prompt. Put that through. Whoops. I actually put that one in twice. First up, we've got I met a real alien in the woods. Glorp. There he is. He kind of looks disgruntled again. It does just sort of look like a man in the suit. Kind of like how the other one looked like a statue. It seems to be having a difficult time making like an alien that looks like it's from a movie or something like that. But to be fair, our prompt here is pov. I met a real alien in the woods. His name is Glorp. So honestly guys, I wouldn't be able to tell that this isn't real. Video footage. It's. It looks straight up real to me. There isn't really any telltale signs. It looks like a guy in a big alien costume standing in the middle of the woods. There's not a ton of movement though, to be fair, so it's not really pushing the model as far as it can go. Next up we've got the beautiful wedding ceremony between Shrek and Bigfoot. And this one definitely was having a little bit more of an issue here, but it still looks pretty cool. I really like these sweeping immediate growing plants across the circle, which kind of makes this a beautiful romantic moment. You know, they're taking pictures of Shrek and Bigfoot, obviously. The clear issue here is that we're getting subject confusion, doesn't know which one to make. Shrek. It looks like Shrek is both of them, but this is more Bigfoot than this. So I'm assuming that the female here is Shrek and Bigfoot is just a Bigfoot with like a Shrek mask on. It's like a Bigfoot Shrek hybrid. So there you go. But the people look so realistic. It's very interesting the way that the model is trying to combine different concepts together and mash them into one somewhat coherent video. I like that they're holding hands there and then the video cuts and then they're no longer holding hands and they're just taking photos and videos. Well, this Shrek is holding the flowers. It definitely looks like a real wedding. Alright, here is our close up macro footage of a mushroom village. And here it is. Okay, very interesting. The model seems to lean more towards realism, especially in this scenario and probably some of the other scenarios we've seen so far. Tiny insects filled between the mushrooms, their movements quick and lively. Yeah, Ladybug crawling up the stem. Okay, so it's trying to capture the miniature ecosystem that the prompt describes. It's definitely not a mushroom village though. It sort of just decided to ignore that piece and make, you know, some mushrooms with ladybugs kind of crawling all over them. And I mean, this looks a little gross, but it's also kind of cool. It's very intriguing to me how well the model is able to keep all of these legs and movements of the legs consistent. I mean, there's a little bit of mushing, there's a little bit of warping, but overall you can, at least while the video is playing, distinguish the individual legs and it's not super jarring. And also, you know, lots of detail here, plenty of detail on the mushrooms. Super close up macro shot. I Love the reflections on the ladybugs as well. Some of you might be grossed out with this and I'm sorry for that, but it's not a half bad generation. It's pretty impressive. Home VHS video Christmas opening a present to reveal a mini nuclear reactor inside which is glowing green and radiating. Oh, lovely. Let's generate that. Here's the tech reviewer video. All right. Definitely your average tech reviewer with the background and the computer, but doesn't look like he's throwing the computer into a thing of lava. It's interesting because as I'm generating these videos, I'm learning more and more about how you should prompt VO2. It definitely doesn't want to be prompt like a YouTube video here. I mean, it definitely will make something that's like a Tech Reviewer YouTube video, but it's not going to capture the part where it's, you know, actually throwing the gaming computer into a volcano. You want to be very descriptive, I think, and very literate with your prompts. I want to retry that one. Wide angle view POV at the edge of a active volcano, we see a man huck a brand new RGB gaming PC into the volcano. Yes, alright, that's a little bit more literate and descriptive. Cinematic movie footage tracking a woman with an umbrell umbrella in New York City. She looks mildly displeased. Lemons are plummeting to the ground from the sky, bouncing off her umbrella. See, I'm trying to get those lemons to have a little bit of physics there. And we've got another mushroom village prompt here that went through and finally generated. And it's pretty much exactly the same as the last one. They're extremely consistent, very difficult to tell apart here. But again, you know, we see the ladybugs are sort of crawling all over the mushrooms. It seems like this model really does, you know, focus on realism here. We're gonna try this same exact mushroom prompt in Sora. Oh, and our VHS Christmas mini nuclear reactor is here as well. And you can see definitely does not look like VHS footage, which is really interesting. I prompted home VHS video, but it seems to be more like, you know, a YouTube video of some guy exploring this mini nuclear reactor instead. It's definitely still Christmas time. I mean, he's weari the Christmas hat, which is pretty cool. We've got the Christmas tree here and it looks very realistic. And I do really like the mini nuclear reactor as well, but it's not home VHS video, which is so interesting to me. It seems like the fine tuning that's Gone on with Google VO2 is preventing it from accessing certain artistic qualities, shall we say again, seems very much posed for realism. I wonder if they're working on some other sub models in the background that are focused on different tasks. Maybe more focused on animation or focused on different styles of video. Not just things that look like YouTube videos, I suppose. Alright, now we've got the gaming computer going into the volcano here. And again, you know, this comes so close. This comes so close. It's only five seconds long. So maybe if it was a 10 second generation we'd actually see it getting thrown into the volcano. But yeah, it's just sort of some guy here holding this gaming computer. He's ready to throw it in, but he's not doing it just yet. I gotta say though, looks like a pretty real realistic gaming computer. You got the power supply down here, CPU cooler, looks like there's some ram. We got the fans and the gpu. It looks pretty solid. And I like this video. Like I like the way that he's holding and manipulating the object and everything. And I like how realistic it looks. It's just he's not throwing it into the volcano. Let's retry this one and we'll also enhance it. Maybe the enhancement will work wonders. You know what I gotta say, OpenAI Sora is also just sort of generating mushrooms with ladybugs. It looks like this prompt maybe is just too long. Honestly, I definitely prefer the Google VO2 result. Although these camera movements are pretty cool. Like the legs on the ladybug don't look nearly as good as it crawls on the mushroom. This one again is pretty cool, but we only see one ladybug here and it doesn't look very nice. It looks a little deformed almost. Let's see if we can get this gaming computer into the volcano. And you know what, let's try the home VHS video as well in Sora. Little comparison. Well, we've got the disgruntled woman as well with the lemons, but they don't really seem to be falling from the sky. It looks more like we've got this video of a woman with an umbrella. And then the lemons are like an animation that was pasted over the video. Very realistic still. And definitely lemons falling from the sky, but they're not bouncing off of her umbrella like I requested in the prompt. This is the volcano gaming computer prompt. Like we've got the guy here at the edge of the volcano and then we've got, you know, something that looks kind of Like a gaming computer, but it's sort of just being, like, lowered down here. And then we've got someone's legs. It's like, literally someone's holding it with their toes, and they're being held up by a crane or a helicopter, maybe. A very, very silly video. It's kind of funny, though, honestly. All right, now we've got another guy. Oh, there's another gaming computer kind of going into the volcano here. It's having a hard time stringing these concepts together, for sure. See, this is what I'm saying. VO2 is more coherent in this sense in comparison to Sora. Like, these are very obviously AI generated. That doesn't look like a normal gaming computer. Google VO2, though, definitely comes out here with something that looks a lot more realistic here. Let's see. Do we get the gaming. Oh, my God, he's going into the lava himself. He's trying to throw it in very clearly. Again, AI generated, but it comes a lot closer. It looks like he's trying to get rid of it, but somehow the universe is holding that will against it, saying, like, no, the computer must not go in the lava no matter what. Oh, wow. Interestingly, Sora does a lot better of a job trying to capture the VHS footage with the mini nuclear reactor. Hey, this video, the kids looks very realistic, though. Good job, Sora, for that. But, yeah, definitely kind of looks like.