Move Over Sora: Veo2 Is The New AI Everyone Wants

The videos you're seeing on the screen right now, some of them are real and some of them are AI generated. I saw Marques do this in one of his recent videos about Sora, and I feel like even since Sora's come out, we've leveled up even more with AI video generation. Now, people that are really paying close attention can still probably tell the difference between what's real and what's not. But now that Google has started to roll out access to their new VEO Video generator platform, we're truly getting to a point where it's more and more difficult to distinguish between reality and and AI. Now, I was one of the people that was fortunate enough to get early access to this new VEO platform inside of Google Labs Video Effects. But if you're interested in playing with this, you can go to this URL Labs, Google/FX Tools video effects, and there's a waiting list that you can join. Ideally, if you join that waiting list, you'll have access fairly soon as well. Once you're logged into Video Effects, you actually have two options. You have a text to video and a text to image to video option. Now, please keep in mind that I did get early access. So by the time this rolls out, there might be more settings, the layout may have changed a little bit, and there may be more or possibly less guard rails on the videos that you generate with it. To get a video, I can give it a prompt. I like to go more detailed, so I gave it a cinematic high action tracking shot follows an incredibly cute dachshund wearing swimming goggles as it leaps into a crystal clear pool. The camera plunges underwater with the dog capturing the joyful moment of submersion and the ensuring flurry of paddling with, with adorable little paws. Sunlight filters through the water, illuminating dachshund, sleek wet fur, et cetera, et cetera. I find it helpful to start with a basic prompt and then use something like ChatGPT or Claude or Gemini to improve the prompt and add more details. I can now click Create Videos and after roughly three minutes it gives you four videos. So we got this one here, which looks more like a dog jumping out and back into the water, but it still looks pretty realistic, honestly. We got this one of a dog underwater and it's kind of showing it more coming out of the water. So it sort of reversed my instructions, but again you can see the realism is there. Here's another one with the dog kind of popping out of the water as opposed to coming into the water. And once again, Another one of a dachshund wearing goggles swimming underwater. It sort of messed up the order of things that I asked for in my prompt. It was supposed to leap into a crystal clear pool while the camera plunges underwater with the dog. But I, I seem to have gotten a lot of videos that are kind of the reverse. Then we have the text to image to video model. So this is like image to video that we've seen. But right now it looks like you can't upload your own images. You have to essentially generate an image with Google's image in first and then turn that into a video. I'm gonna go ahead and do the I'm feeling lucky and we get a cinematic moody shot captured a fair skinned man in dark leather jacket leaning against a silver car in the foreground. Let's just go ahead and see what happens when I try to get lucky. We get this starting shot here, which looks pretty decent. And then I can move on to step two, which is the motion step and we can describe the motion we want to actually see in the video. I'm going to go ahead and use the prompt that it sort of pre filled in for me because it's a pretty good one. A cinematic short film opens with a low light shot of a man in a leather jacket leaning against a vintage car, et cetera. Let's see how it animates it this way. Let's go go ahead and create the video once again. It's going to take a couple minutes and then we should get four outputs from it. And after about five minutes we got these four outputs here that all look pretty good. I mean they look like a cinematic shot that you would see from like the opening of a TV show or a movie or something like that. I'm pretty excited about it now. If we like what came out here, but we wanna see more, we have the option to get more variations and once again it'll jump back to our original image and start to generate more variat. But I'm gonna go ahead and hold off on that because I took to the social medias, I went to LinkedIn, I went to Threads and I went to X and I asked you to give me some prompts that you'd like to see me generate in VEO. I received over a hundred comments on X plus a handful more on LinkedIn and Threads. So I wasn't able to generate every single one. But I did manage to test about 30 of the prompts that came across and again, all 30 of them came with four generations. So I don't even know if I'm gonna get around to showing every single thing I generated. But I will let you know, these were all the first prompt, I did not do any cherry picking. I took the prompt that you suggested, I plugged it into Video Effects here, and then I downloaded the four videos that it gave back. So there is no cherry picking here. This is what they gave me. You're gonna see the good, the bad, the ugly, all of it. One thing that this Video Effects platform does sort of lack at the moment is it doesn't save the previous prompts that I made. I can't go back through and look through, like, a library of videos that I made. So instead, I downloaded every single generation that I made based on your prompts. I actually even recorded myself entering the prompts and showing the four generations on a screen like this. But actually showing you all of this video playback of me entering the prompts and showing the results seems like it would be a bit of a waste of your time. So instead, I'm just going to show you the outputs that I came up with. And you're going to have to trust me that I didn't specifically cherry pick any of them. I actually entered the prompt and downloaded the four videos. And just for fun, as a comparison, every single prompt that I entered into VEO I also tossed into Sora. So we can do a quick side by side comparison of how well SORA did it versus how well VEO did it. So let's start with a simple prompt here. My buddy Michael Stelzner here said, show a dog surfing in La Jolla. Here's the first video it generated where it kind of looks like the dog is running on the surfboard. So not the best generation. And then the surfboard kind of disappears. Here's the second generation. This one kind of looks like there's a GoPro or something stuck to the of the board. Looking at the dog. This one came out pretty impressive, in my opinion. This one was probably my favorite generation of the 4. It looks like a sort of collie on a surfboard. And it looks real. I mean, it looks like a drone shot sort of hovering over the dog. And this actually kind of looks like La Jolla in the background. So that one, to me, came out pretty good. Finally, we have this one, which looks like another sort of GoPro shot, but the dog kind of changes age. If you notice, at the very beginning, the. The dog looks like a little puppy. And then as the video goes on, the dog stands up and looks like an older dog. So there's a little bit of like a coherency issue there, but still pretty good. And to compare it to what Sora generated, here's my two dog surfing images that Sora made. One, the one on the left, not great. Actually, neither of them are really that great. I think almost every single one that Vaio did was better. Next up, Barnabas Lawson here suggested three. I actually tested a group of of people parachute jumping out of a plane. Here's the first generation that came out of it. You don't really see the plane and all of these people seem awfully close together, but that's what it came up with. Here's the second one. It's like a tandem jump. There's an extra arm going on here, I think. 1, 2, 3, 4, 5. I'm counting five arms, but yeah, so that's the second generation, not the best. Here's our third generation. The selfie stick here just kind of appeared out of nowhere. But I do like that angle where it put the selfie stick out and you can see it. But it's just, it's weird how it sort of appeared out of nowhere. And then here's the final generation where two people are strapped to one person on like a three way tandem parachute jump. So if we take a look at Sora here, these are the two that Sora generated. The one on the left, it kind of looks like they're just hovering near the plane. The one on the right, there's two people that somehow eventually merge into one. Neither of them actually really generated great videos with this prompt. Honestly, Matthew Schrader asked to do Will Smith eating spaghetti underwater. Unfortunately, fortunately, VEO will not let you generate real names. So if I try to do Will Smith eating spaghetti and then create the video, we'll probably get an error. Yeah, we couldn't create what you asked for. Try a different prompt. So whenever you try to put somebody's real name in there, it doesn't work. Sora, on the other hand, did let me enter Will Smith eating spaghetti underwater, but none of these videos look anything like Will Smith. So it's a guy eating spaghetti underwater, but I can't really call that Will Smith. Chris Johnson here said a hamster the size of a skyscraper dressed as a samurai battles a swarm of tiny flying drones carrying paintball guns. I liked that prompt. Here was the first video it generated, which kind of looks more like a still image that has some like parallax effects going on. Here's the second video once again, sort of looks Like a still image with some paintball animations going on in the background. Here's the third one. This one's probably the best looking one where you can see the drones kind of coming over the shoulders. That looks like a very kind of cool, cinematic shot. Probably the best of the group. And then finally it gave us this one where once again, the hamster looks totally frozen and the drones are just kind of going around in circles here, you know, spitting paintballs. So one out of the four wasn't too bad. Comparing the same prompt over on Sora, here's what we got. Once again, the one on the left looks like a static image that just sort of zooms out the one on the right a little bit more of kind of what we were looking for. Sora probably actually kind of wins this one. Now, one thing to note is that with Sora, I was generating in 480p, so do your best to sort of ignore the pixelation there, because I was trying to get fast generations, not the highest resolution generations. Moving over to threads here. Nathan said a dolphin is breaching from a sea of honey. I was curious what that would look like. Here's video one of a dolphin breaching from a sea of honey. You don't really see the breach. It almost just looks like a dolphin swimming in some, like, yellow water. Here's the second generation. Once again, it kind of looks like a dolphin swimming in some yellow water. And right before it looks like it's about to breach, the video starts over. Here's the third one where we actually see the dolphin breaching what looks like a sea of honey. You can actually see it's got that sort of, like, honey viscosity. But as the dolphin lands, it feels like it's in the air for a little too long. Like it's kind of hovering there a little too much. But that does look like that sort of honey viscosity. And it does look like a dolphin jumping out of it. And then finally, here's another dolphin breaching the sea of honey. Once again, this one, to me, just kind of looks like yellow water. And there's some weird bubbles there. One out of the four looks pretty good. Comparing this to Sora, here are the two videos that I got out of Sora. Yeah, they don't look great. The one on the right almost looks like the dolphins turning into honey. I'm jumping over to X now. This is where I got the most requests, and I apologize if I wasn't able to actually get to your request. Here's one that I actually posted just in one of my replies to X where the prompt was Mysterious young woman conjures Lamora, her hands enveloped in swirling translucent flames. Fed by her raw emotions, the crimson fire twists and turns, illuminating her sharp features and casting shadows of her inner turmoil. And this right here is what was generated, and I think it follows the prompt. Pretty good, honestly. This one was interesting to me. A slow melancholic pan across a dimly lit DVD rental store. The camera focuses on a middle aged customer wearing casual attire, his face esch with disappointment. He sighs deeply, then gently places a DVD case on a shelf with reverence. As he steps back, the camera smoothly zooms in on the dvd, revealing the title Squid Game in bold neon red lettering. The case sits nestled between Squid Squid Game 2 and Squid Game 3, their spines glowing faintly in the store's fluorescent lighting. The shot lingers, emphasizing the irony of the original's placement among its sequels. So the generations you just saw were what Veo did. And here's what Sora managed to generate with that exact same prompt. Definitely think Veo did it better. Steven Greenwald asked for A man is playing with his dog at the beach. In the background, dolphins are surfing the waves. He goes on to say, most leading tools have issues with this. And as we can see from some of these video, generations here also seems to have issues with this. Here's another one where the dolphins just seem to be like frozen in midair. I mean, you got the guy playing with his dog at least. Here's another one. The guy throws the ball up in the air and then it cuts to a shot of dolphins with a ball floating above them. And here's the final one. A guy petting his dog. And then some just like frozen dolphin images in the back. So yes, Vao struggles with this one as well. Somebody said to give it the prompt AGI, so I was curious what that would look like. So right now on the screen, I'm showing you the four videos that it generated when I gave it the prompt AGI. All of them pretty wildly different. I always like giving abstract prompts just to see what comes out of it. When I gave Sora the AGI prompt, these were the two videos that Sora generated for me. Both pretty similar actually. Someone suggested a wolf howling at the moon, which I always like to test. Here's the first video I got of a wolf howling at a moon, which looks pretty good. I don't know about the moon and why it's moving so fast. I guess it could sort of be a camera panning around the wolf. Here's the second video it gave me with like a weird red moon. And the video literally just freezes halfway through. So this one got screwed up for some reason. Here's the third wolf howling at a moon here, which looks pretty good, but the background is red and you can't really see the moon. And here's the fourth video, which looks really realistic to me. But once again, the moon isn't really in the shot. But compared to Sora, giving Sora that same really basic prompt, it just gave me like a frozen picture with like a parallax effect on it. Gabo here gave me this long prompt. I'm not going to read the whole thing, but basically it two plant seedlings growing taller together and very close by. I would say Veo kind of struggled with this one. We can see some plants blooming here, but yeah, they're a little bit funky. Here's another one showing the plants blooming, but not much else. So it sort of ignored a lot of the other details that were put into the prompt. Here's the third video where the plant, you can't even see it cut it off at the top, but you can kind of see some like movement here in the soil. And then finally it flashes a flower at the end. And then this one I actually thought was cool looking because it kind of looks like after effects animation showing two plants growing. But it also really didn't follow the prompt very well because there's nothing really growing. And here's that same prompt when given to Sora, also not great. So neither really did that one too well. Colton Batts here said a vaporware styled cinematic wide shot of a robot detective under a streetlight. And it's raining. I actually think these generations came out pretty decent. I've got all four of them up on the screen right now, so you can compare them. And here's Sora's version of that same exact prompt. Once again, I definitely think Vaio wins in that competition there. Ahmed Reza here says Bollywood dance sequence in Alaska featuring grizzlies and polar bears. I've got all four of those up on the screen right now. I actually think these came out really, really good. I think some of the faces don't look amazing, but I was pretty impressed with how this came out in general. And when you compare it to Sora, here's what Sora gave us again. The 480p that I generated in isn't really doing us any favors, but not horrible as well. There's even less detail in the faces. But I think both of them did this prompt pretty decently, if I'm honest. This was a really fun one from Jake Dan here. A punk rock squirrel in a studded leather jacket shouting into a microphone while standing on a stump. This one was fun and I think a few of these actually look pretty solid. And here's our comparison of Sora giving that same one. The one on the left almost looks like a little like animatronic squirrel or something. And the one on the right, the squirrels just like eating the microphone. Still pretty funny videos though. Andre suggested create a rough seas and stormy weather lighthouse in the middle of the sea, Big tall tsunami waves cloudy and galaxy sky. Realistic, minimalist, surreal. So that's what you're seeing on the screen right now. I'm just putting all four of them up. And here is that exact same prompt when given to Sora. I actually think the one on the right for Sora doesn't look too bad, but I also think that Veo definitely nailed it a little bit better than Sora did. Mikolalak here says a spider Kaiju sipping tea with a dragon Kaiju over the burning remains of a city they just destroyed. Now, the spider part wasn't too great, but it seems to have gotten the Kaiju dragon. And it also seems to have gotten the destruction part. Comparing that to Sora, here's what we got. It got the spider, but it didn't really get the dragon portion. It also did get the city destruction pretty good. Illuvium Gamer said cat riding a unicorn through space. And what you're seeing right now is the cat riding a unicorn through space from Vaio. They're okay, not too bad, but let's go ahead and compare that to what Sora gave us. I actually really like the one that Sora generated on the right. I think that's kind of hilarious. It also didn't give the unicorn wings where Veyo, for whatever reason, gave the unicorn wings. Restructuring democracy suggested a cat wearing a cat suit, eating a cat in its mouth, standing in a cat in a room decorated in cat pictures and full of cats eating dogs. I don't know if it really nailed the last part about the cat eating dogs, but this is what you get from it. And here's our Sora version for comparison. Once again, no cats eating dogs, but pretty silly little animation nonetheless. After Light asked for a person solving a Rubik's Cube, it really seems to struggle with the Rubik's Cube motion and the colors actually changing properly. So wasn't Super Impressed by what it managed to generate with this prompt. But comparing it to Sora, I think it actually did a little bit better than Sora because on Sora, one guy is just touching it and the colors are changing. Where Veo actually looked like they were doing the motion properly. Gooey here said two astronauts playing ping pong on the moon while Earth explodes in the background. Well, we've got the astronauts, we've got a ping pong ball. It doesn't really look like they're interacting with it properly. And you can definitely see Earth exploding in the background. And here it is. Compared to what Sora generated, Sora has less of the explosion in the background and they actually look a little more like astronauts. They're actually wearing the proper spacesuits and everything. Neither of them really nailed this one, let's be honest. I also saw this prompt from my buddy Belavel Sidhu. A side scrolling shot of a rhinoceros walking through a dry field of low grass plants. And you can see his comparison with Veyo and Sora here. And then I went ahead and did my own version of that and got this one and thought this was like the most realistic AI generated rhino I've ever seen. It is really, really good. Here's some other footage that I tested. These weren't from actual recommendations. They were just random prompts that I tested for myself. CCTV footage of an alien walking in front of a convenience store. And we can see the four videos that it generated there. None of them really scream CCTV footage to me, especially since the camera is moving, but still pretty cool. Here's a first person view out the window of a jet fighter. I thought all these looked really, really good. The one on the top right I thought looked especially good. So impressed with that. Found footage in a dark forest with mysterious monster. And for whatever reason it only generated three out of the four and didn't generate the fourth one. I really like the found footage up on the top right. I thought that one came out really good. And here's one of the image to video videos that I made with the circus tent here. And once again, it only generated two out of four. I don't know why, but the two that it generated were both, both pretty impressive looking, like establishing shots that could easily be used for like a movie about the circus. So there you have it. There's a sort of deep dive into Veo and what it's capable of. Again, I tried to test as many prompts as I could. We'll probably play with it even more on One of my Monday live streams, if you're not familiar, I am doing live streams every Monday at 11:00am Pacific Time. And if you want to test out Veyo, we can do that on the next live stream. And I will take your prompts live, we'll generate them on that stream, see what we come up with. But just kind of quickly giving my thoughts, I think that we have a new standard for AI Video. I think Veyo is much better than what we're getting out of Sora Turbo right now. We haven't seen the full version of Sora. The Turbo model is a sort of quicker to process model. It doesn't generate the greatest results. So I think OpenAI still has a better Sora model. But of what's available for people to use right now, Veyo is definitely winning. And yes, a lot of the generations you're seeing online are cherry picked. But Veo is kind of designed for cherry picking, right? You give it a prompt and it's going to give you four videos. And everything I showed you was the four videos that it generated. I'd never generated any of these prompts twice. It gives you four videos and then you cherry pick from the four that it generated. It does still tend to struggle with people doing gymnastics and a lot of really complex prompts with things supposed to be happening in the foreground and the background. Like we saw with the guy playing with his dog and the dolphins, it almost like focused on just one or the other. And as we always say, and I know this is cliche, but this is as bad as it's going to get. This is a really, really impressive model. I'm grateful that Google gave me early access so I can demo it and play around with it and show it off to you. But AI Video seems to be jumping by leaps and bounds and it's only a matter of time before this is publicly accessible to everybody. And man, that next version of Sora and that next version of VAIO are probably going to really blow some people's minds because this one is already starting to blow people's minds. So that's what I got for you today. That's my breakdown of Veyo again. We'll probably do more tests on a future live stream, so make sure you join me for one of those. But I really appreciate you tuning in and checking out what this new technology is capable of with me. It's been really, really fun to play with and I'm excited to dive deep into even more tools like this on future videos. So if you're not already subscribed, make sure you subscribe to the channel. So thank you once again for tuning in. I really, really appreciate you. I'll see you in the next video. Bye.