Limitless: An AI Podcast - Video Will Never Be The Same: How Veo3 is Breaking the Internet

Episode Date: June 19, 2025

Google’s new video model Veo 3 just let one creator crank out a $2 k, three-day commercial that ran in the NBA Finals—proof that AI can out-produce multimillion-dollar ad agencies overnight. We break down how its eight-second clips with built-in sound now enable talking characters, viral absurdity, and creator-less channels, from stormtrooper vlogs to fake pharma spots. But Veo 3’s realism also fuels deep-fake missile strikes and reality-bending stunts, so spotting the fraud is becoming a full-time job. Stick around for our chat with solo mastermind PJ Ace to hear how he pulled off the ad that fooled millions.------💫 LIMITLESS | SUBSCRIBE & FOLLOWhttps://limitless.bankless.com/https://x.com/LimitlessFT------TIMESTAMPS00:00 Veo3 Is Out Of Control10:37 Stormtroopers20:26 How Far Does This Go?23:53 Fake Missiles28:59 Pro's And Cons32:21 What To Watch------RESOURCESDavid: https://x.com/trustlessstateJosh: https://x.com/Josh_KaleEjaaz:https://x.com/cryptopunk7213

What if I told you, Josh, it draws that the most viral ad that aired during the NBA finals was not made by any sort of ad agency, but instead by just one person using AI in just three days for just $2,000. That's exactly what happened with V-O-3. We talked about VO-3 on a previous episode. It's Google's brand-new AI video generation model. It dropped 20 days ago. And already, videos out of V-O-3 are pulling millions.

of views on TikTok, YouTube shorts, Instagram Reels. V-O-3 is different. When it got released, it started impacting the internet, basically, immediately. And to prove it, we're going to watch the exact 30-second ad that aired during the NBA finals, produced entirely by one person using V-O-3. And then after, the boys are going to talk about why this changes everything. So let's go watch that ad right now. Indiana going to...

Asking people what they put their money on. Indiana got that dog in them. Will egg prices go up this month? I think we'll hit $20. How many hurricanes do you think we'll have this year? Cali! Cali lets you legally trade on anything, anywhere in the U.S. OKC!

Indeed! OKC! Okay, C. Okay, so Calci is a prediction market's platform. So you can bet on events. And so this ad went live during the NBA finals, of course, to say, like, you can actually bet on what NBA team you think is going to win or lose, but also other things. They hinted at other things. Egg prices, hurricanes, and that's what prediction markets are for.

The cool thing is, is that, of course, as I said, this was just made by one person and Google's V-O-3 model, and it is already doing the job of what would be a full-scale, full-scope ad agency, ad production company. And it was just made by this one individual who's on Twitter. His name's PJ Ace. We're actually going to interview him later today. So that episode will come out on the feed, which is why you should like and subscribe. So when we talk to PJ Ace to get all of his information, you'll be able to hear that. Josh, when you saw this ad go out, what was your first thought? What was your first reaction?

This is incredible. The first reaction was no reaction because it's like, oh, it was just another commercial. And I think that's what was so incredible about it. Did you see this live? Yeah, I was watching the game. I saw this. I was like, wait a second. Like this, you start to form this like reality filter when you're looking at things being

people like us who are always looking at AI content. you're like, wait a second, that kind of triggers some filter, but I'm not quite sure, until it gets a little further in the video. But it looks so good. And I think something like this was not possible three weeks ago even prior to the O3. And it shows, it's a testament to how fast things are moving. I think the fact that he was able to do this with, what, $2,000, one person in a very short period of time, relative to the multi-million dollar budgets and hundreds of people production crews that they did it with, it's amazing. And I think because they were the first mover and because they were the first month to actually deploy an AI ad, it got a ton of attention. And I saw

this all over Twitter for days. I was at dinner with some friends last night and we were talking about this ad. And one of my friends, we were talking about it being AI generated. And he just responded, that was AI, which I think if you're not like looking at it with your entire brain, you will not realize that it's actually AI. There are like obvious giveaways, obvious tells the speaking alien being one of them, right? The absurdity of some of the scenes being one of them. Some of these scenes, an alligator in an inflatable pool. Not going to happen in reality. But if you're not using your entire brain, that detail is just totally going to slip by you. Yeah, can I be honest? If this video wasn't as unhinged as it is, I would have believed it was real. It was about the point in the

video where you get to the guys swimming in the pool of eggs. Weirdly, that could actually be a real thing where I was like, oh, this is probably AI. And now I'm like, noticing the alien is actually not a human in like some kind of like cosmetic makeup or whatever. This is insane. Josh, you said this cost $2,000. Yeah, David, was that that that was the price? That's how much. That's how much PJ Ace. That's how much he said that it cost him to make it.

Yeah. And he did it in two and a half days. That's amazing. So V-O-3, it works eight seconds at a time. So that's really piecing a lot of these together quickly. Downstream of that, there's an entire category of YouTube videos that are more or less all titled how I use Google V-O-3 to create viral videos. And I found like eight of these videos that had over 100,000 views a piece. A few of them had a million views apiece. So you already know

that the creators on YouTube, just people who are creating content, are already looking at how I master V-O-3 to create viral videos. This is a video I watched of a guy just making making a video. There's a guy who made a Sasquatch video. So people are already experimenting with V-O-3 about how to master virality. But you can definitely see that in the Kalshi ad where the absurdity is just viral. The ad itself is absurd. And that's contributing to why people are talking about it. But you can make much more absurd content using AI than you can actually film in real life. I remember when Jackass was a thing I'm probably dating myself. But like Jackass was Johnny Knoxville and all the other people doing absurd things. And it was like 10 minute clips, 15 minutes.

minute clips and they're doing absurd things, but they were truly doing things that were dangerous and unsafe and unhinged. And now you can just create those with AI and go viral due to absurdity that way. Yeah. So it's safer. It's cheaper. And you can basically fit in any of your favorite characters or fan fiction, human or unreal, in these videos. That's just insane. You know what my mind goes to immediately is like, weirdly, like, what real actors and actresses would sell like their IP to put into like these videos, right? You mentioned David just now that like we've seen like Stormtrooper vlog videos and sometimes I saw like a video the other day. Yeah, I saw a video the other day of like someone pretending to be like a biblical character but like vlogging in the

modern day era. I've watched like five of those videos and I follow a few accounts. I'm wondering like where this goes when like actors just kind of like just say, hey, you know what? I'm kind of like lazy. take myself from my heyday, make me look prettier, and then just fit me into like a bunch of these different vlogs. I feel like people can take this anywhere. It's insane. Yeah, Siddy Sweeney will just be a permanent actor at her current age and she will never, ever age. And she'll just be immortalized and AI. I'd be a lot smarter, David, if that was the case. If that was the case for days ago. There's also the case, though, is why limit it to people who actually exist? Because that seems like a lot of royalties, a lot of like chaos. You could just create these new actors and

actresses and characters that are coherent throughout this world and you could create these worlds. And one of the interesting things to me is that, like to the untrained eye, which is most people, like we spend a lot of time looking at AI content, but most people watching the NBA finals do not. There is no difference between this ad and any other ad. And for them, like, oh, maybe there's some CGI that's happening that is traditional. But I think the awareness that this is AI instead of just a traditional CGI or normal commercial, it isn't really there. So you get all of the upside of cost savings, of creativity, of productivity, without any of the downsides of actually getting a lower quality product. And this is the worst it's ever going to be.

And I can imagine by the time we have VO3-4, it will be nearly indistinguishable from reality. So it's like a hugely powerful unlock. Let's watch some more content that has come out of V-O-3. So I got two more ads to show you. Both of them are, one of them is not a real ad. One of them is satire. One of them is real. And so let's start with the satire one first. This one was just made as an example of what somebody could do. And actually, let's just watch this. Let's just watch this fake pharmaceutical ad.

And it does a very good job of simulating what could be. Let's go watch this. I tried everything from my depression. Nothing worked. Every day felt heavy. I felt trapped. Then I tried Pupperman. Our prescription helps your body secrete a special pheromone that attracts puppies.

I took the pill before bed, and when I woke up, there he was, the love of my life. The pill does not target depression directly, but we've found that it's really difficult to be depressed when cute dogs show up at your doorstep. I used to feel so empty, but now I feel joy and mild concern how a pee stain got on the ceiling. All right, we don't have to watch the whole thing, but, you know, satire ad, Pupperman, it's as far as pharmaceutical companies go, like, ad ad, ad, go, which are already so contrived, an AI alternative. Actually, I feel like slots in very, very well. Now, you can pick up some patterns, I think, between the call sheet ad that we watch, the fake Pupperman ad that we watch. It's a series of clips. And so it's a little bit schizophrenic. It moves from scene to scene. The clips are maximum eight seconds long, because that's just

what Google V-O-3 does. So there are some definite tells. And so I think while we are talking about this as if it's brand new because it is brand new, I think it's going to become very easy to tell as time moves forward what is a video made by Google V-O-3 versus what is not. But again, those tells are just, you know, the first iteration of this thing. I don't expect those things, those tells to actually stay put for very long. It is probably also important to explain how V-O-3 is so much better. Like, why suddenly do we have this giant jump? And the big breakthrough with V-O-3 was the audio. We had pretty good video generation. prior, but you couldn't really have a character speaking to the camera. There was no perceived

physics engine in the video generation. But with V-O-3, the big breakthrough was like, you could have people that talk. And a lot of the early viral videos were just characters talking, and they have their own voice. And if they're in an echoy room, you hear the echo. And there's this new category of ASMR where you can slice concrete with a knife or like you can just do these weird. Yeah, but it emulates what it would sound like. And that was a huge breakthrough with V-O-3. And the reason we're seeing all these videos go viral in a way that they never have before is because now you have the audio component and it's really good. And I mean, I'm sure the eight second unlock will come where it will be longer, but you can get a really good quality, very convincing eight second

clip that sounds real, that looks real. And that's the real breakthrough with V03 was the audio this time. I've got to say, one of my favorite parts of this trend for VO3 are the vlogs that I mentioned earlier. So take a look at this example of this is a stormtrooper, which is a very popular character from Star Wars. And typically, Stormtroopers get, you know, for all the nerds out there, they get a bad rap. You know, they're the evil guys. They're kind of following the kind of like Dark Lord or whatever. You can tell that I haven't watched too many Star Wars episodes. But typically they're the first character to get integrated into any kind of fan fiction or novel for people that love Star Wars, right? People dress up in Stormtrooper outfits and all that kind of things. And

what is not commonly told is the story of. of these different characters. And what someone had the bright idea of doing as soon as V-O-3, you know, this is a very important AI tool, as Josh just mentioned. It integrates audio with AI. Imagine all the amazing things you can use this for, you know, education, product placement and stuff. They were like, you know what?

I'm going to make Stormtroopers real. And what's really interesting, if I was to summarize this in a sentence, it's like the early days of YouTube. I think Josh, you said this on a conversation offline. Like, it reminds me of the early YouTube videos where people were like, I don't really know what to film. I don't really know anything about hooks or like main plot lines. I'm just going to video stuff. I'm pretty sure the first YouTube video that was ever made by the founders themselves was of them at a zoo, right? And they were vlogging and they were like, you know, hey, check out this

animal and all this kind of stuff. And it reminds me of the early days of then. And what I love about this is it doesn't need to be kind of like mass mainstream. It could just be personal and unique to you. So you as a fan, as a 12-year-old, 15-year-old or whatever year-old can just be like, you know what? I've always thought of Stormtroopers of doing this, A, B, and C. Like, let's just see what they would do. What would they say if they were in our times today, if they were experiencing all the kinds of things that we were on the internet today? I think that's pretty awesome. And so the creator of, there's an Instagram account now called Storm Trooper Vlogs.

It's got 330,000 followers already. It was made somewhat recently, like in the last two years. two or three weeks because downstream of V-O-3. And then each one of these videos is like 30 seconds long. And it's, again, stitched together of the eight-second videos that come out of Google's V-O-3. And the creator of them is doing a pretty good job making a story arc to them. And so inside of every video, there's a beginning, middle end. But then you can also, like, watch them sequentially.

And you see these Stormtroopers being, like, they're complaining about, oh, Vader has just deployed us. We've got to go to Tatooine. The next vlog, they're on Tatooine. crash landed. And so there's actually like a story arc here. And so whoever is creating these people behind the scenes here is actually, you know, squeezing their creative juices and saying like, hey, let me use this Storm Trooper motif and create a story. I don't know where it goes, but they're making a story happen. Sorry, can we also just appreciate how high quality these videos are? It looks incredible. I don't think even the early Star Wars films could look as good as

these, right? Definitely not. Yeah. What's the name of the show with Pedro Pascal? The Mandalorian? The Mandalorian. Like, this looks as high fidelity as the screenshots as the shots that they have in that film. And I'm just thinking, like, Lucas Arts, which is like the studio that was behind Star Wars that came up with all these different animations of like the Stormtroopers, etc. They were acquired for how much by Disney. I'm pretty sure it was to the tune of like $1.5 billion. And now that moat has been reduced to like what? Zero or like 500 bucks to create this clip. It's just insane.

So Google V-O-3 does all the image generation and sound generation. But the scripting is done by the creator. And you can see people on Instagram like commenting what they're funny the lines that they like. And so Greg got kicked out of the bar for being Greg, you know, smiley face, laughy face, 4,600 likes. So people are relating to this content as they relate to any of. other content that they see on Instagram. The engagement is insane as well, guys. Like, we're talking hundreds of thousands and millions of likes per video.

And these videos are being, like, how many is being released by this account that you follow, David? There's like, two and eight. Yeah, there's already look something like 40 or so. 28 at the top. You see the profile? Yeah, the first one has 10 million views. Yeah. And there's like two parts.

There's the novelty of it where it's like, whoa, stormtroopers now have this advanced backstory. but then there's also the interactive component to it where you can see your favorite character and also kind of have input on the storyline and watch it unfold in real time where the comments are happening and people are kind of saying what they want to see and maybe that's impacting the storyline

and it's this like very real time process of something that would previously take like a couple years to get from one story to the next. Like you're watching one story to the next, this is one every couple of hours and it's being generated in real time and it's kind of this interactive thing and you're watching these characters that you love kind of evolve,

but on this weird unhinged story that would never sell as a LucasArts film, but you can do it because it's just Instagram. So you're seeing the funny, unhinged, like, X-rated side almost of these characters that you know and love it. And that's like a super powerful thing. And you have to ask, like, where does that line get drawn for intellectual property, if anything? Because now you can take the likeness of these amazing characters and build an entire

entire universe on top of them in like an hour. And it looks amazing. I mean, the end game here is that they're getting tens of millions of views. So this has to be monetized somehow. I don't know whether legally or illegally, but like they have to figure out the IP laws or issues with this. Because if I was Disney looking at this, I mean, this is just great PR for the entire franchise. Now, to my knowledge, no one's made kind of any kind of like an illicit version of this. I hope so.

I mean, they are swearing. They're saying naughty words. Okay, right. So they're saying Nottiewood. So it's not PG-rated for the typical audience at Disney. And Greg is getting drunk at the bar. Yeah, Greg's getting drunk. Oh, yeah, so it's worth noting that there are two stormtroopers,

and they are persistent. The identities are persistent. And so one of them is named Greg, and the other one is the main character. I don't know what his name is, but his buddy is Greg. And so it's him and Greg as two stormtroopers in every single video. So even though they're just normal-ass stormtroopers, they're the same two characters over and over and over again. I'm in all looking at the surfing physics.

It's really good. Like he goes off a wave and he falls as if he were really falling. There were some examples of physics not being so great. I can't find the one with a monster and like a Jedi comes in. But then you can see like, oh, the physics on that one wasn't really perfect. So a question for you both, beyond TikTok Instagram entertainment, what do you think V-O-3, with all its audio and video capabilities, could be useful? Like outside of that.

David, I know you mentioned products and maybe you can kind of like pitch your product in, I don't know, a million different colors or using like, really aesthetic characters and, you know, any kind of creative brain can go with that. But is there anything else that we're missing that might be kind of obvious? And maybe it's just entertainment to start off with. Well, I think the way that it goes viral is absolutely entertainment. And going back to, like, again, the early days of YouTube, what did YouTube do? It unlocked much more creators than what previously existed.

Like with YouTube, you just needed a digital video camera. And then all of a sudden you could become a YouTube creator. Like your person could become a YouTube creator. But in this and age, being a YouTube creator is pretty competitive, and you kind of need to be a pretty charismatic orator of good information. You need to be really presentable. You need to put in a lot of work. And if you have just a really rough accent or you're just shy and you're not outgoing enough, it kind of bars you from being a good creator on YouTube because you need to have that like MKBHD style of art in order to be able to be good at being a YouTuber. You have to be a YouTuber.

With V-O-3, you can still, like, actually have a lot of creative juice, creative energy, but you can use V-O-3 as the actual cosmetic overlay over whatever your creative is. And so there's this notion of just a creator-less YouTube account. What's the faceless channel is like the words that I think are going around. It's like, you, the creator, can be a creator, but it doesn't have to be you anymore. It actually just has to be you and your AI. assistant, your V-O-3 assistant to create a story. And if you have the creative juice, there's basically nothing getting in the way between you and a very successful creator channel either on YouTube or

on TikTok or on Instagram. And so I think we're just doing another step function change of another 10x in like people who are good at creating content. You have these like tools of infinite leverage where you're really just limited to your creativity. And it's just like back to your point. If you view video as a tool and you just think of where are we consuming video? it's like, well, there's two main buckets. There's like, we're normally consuming to entertain or we're consuming to educate. And I think it really empowers people in both of those sectors to use these infinite leverage tools that are only limited by creativity to get creative and unlock all these huge things across entertainment, across the board. Anytime you interface with the video to be entertained, that's going to look totally different.

And the same with education where now you have this hyper customized version to teach you anything. If you want to learn how to change something in your sink or your toilet, like you can, you can, you can, just generate a video on the spot. And that's like a new unlock too. Yeah, I mean, my history lessons would have been way more entertaining if I could actually watch the guy talk and tell me about his time living in what the middle ages or whatever that might be, right? I was just thinking, like, what's the natural Pareto effect of this new tool now, right? So let's extrapolate out to like, I don't know, a decade from now. We have infinite more content than we could have ever, our pea brain could have ever consumed or hope to consume, right? And I'm guessing like, okay, Duma hat on, 80% of this

is going to just be slop, right? Dopamine kind of fueled like, ha-ha, look at this ASMR thing of this knife cutting a glass strawberry, but it sounds so good and this is so visually weird and I can't look away even though it's just eight seconds, which, by the way, that is me right now, my entire feed, I would say about 40% of it is just AI stuff. And then the 20% is like real creatives, as you said, Josh, that are using this for much bigger purposes, which will have, you know, massively outsized impacts on education, on advertising, on product placement, and all those kinds of things. Do you guys generally agree with that trend? Or do you think it might be something a little more even or perhaps even the inverse correlation?

It feels like it might even go further than that. Like, I think of my day-to-day workflow. And if people watch the show, they may have seen an original short that I've published where I'm just kind of talking to a camera about a specific topic. And if I can, take myself and place myself in this video generation model and have it iterate and create these original shorts about a topic and do it for me and do it precisely and do it personalized for every single person. I much prefer that reality than meaning to do it to myself because it gets the point of cross, but it doesn't require all of the time and effort and also the lack of customization per person. And I think if you extend that out elsewhere and you could just emulate a person

or create a new one without the rules of traditional algorithms and physics applying, it just creates these much more hyper-customized versions of what we're already used to consuming. So your distribution is probably right where like 80% of it is probably going to be crappy. And I think I do agree with that. But that's already 80% of crap YouTube content. I think it might honestly even be higher than 80%. And I think that's okay. I think it just probably becomes a little more elevated.

But the 10%, the 5% of good content will probably be exceptionally great. Like, it will be better than, and as you mentioned reliving a historical event, I saw one of these videos of the people who snuck into the castle in a Trojan horse. And it was playing out the entire sequence of events as if you were there. And there's this, like, Knight who's got a vlog camera and he's like filming around the boys of the campfire. And he's explaining the process of coming to the conclusion, like, oh, we. We need a Trojan hoist. Oh, we're going to go in it. Here's how we're building it.

Here's what happened when we got in. And it was entirely historically accurate, I'm sure. Exactly. And you could really, you can like experience the mid-century of like wars as if they were a vlog on YouTube. And a very high quality one that's historically accurate and contextually accurate. And that is like such a powerful unlock. Like, man, if I had that in school, that would have changed everything.

Dude, I couldn't agree more. But David, you make a really interesting point, which is, okay, what if this tool gets in the hands of people that want to kind of drive a different kind of agenda and maybe pervay certain historical inaccuracies or current inaccuracies on current topics and political stuff. Do you have a take on this? Yeah. So this is perfect because this is exactly where I was going to go next. There were a bunch of, due to current events with the Israel-Iran conflict. There were a bunch of fake AI videos of Iranian missiles landing in Tel Aviv that did not actually happen. The explosions themselves were AI generated. Now, I saw these on Twitter. I'm on Twitter

all the time, and so I saw them go out on Twitter. They have since gotten deleted, which is interesting. I wonder if that was Twitter doing that. But nonetheless, for a solid 48 hours, there were seemingly very real. You could kind of tell if you were, again, if you were paying attention. But, I mean, I think we here are as good as you can be for reasonably understanding what's AI versus not AI, average population will not. And there was definitely a campaign coming out of these like unofficial Iranian Twitter accounts that were tweeting out videos where not real missiles were landing in Tel Aviv. And I actually had to like send some of them to my friend in Tel Aviv, being like, hey, I'm pretty sure this is AI, but like double check for me. And she like kind

gave me the affirmative versus not. And so again, this is just, you know, week three of Google V-O-3. I don't even know if it was Google V-O-3, but the conversation is the same. You can definitely see how this could, like, turn into big disinformation, misinformation, misinformation campaigns to create a narrative. It's a super powerful tool because, I mean, going back to the first example with the CalShe ad, I would bet over 80% of the people watching that didn't even recognize that it was AI. And it's gotten so good now where you can really just reshape reality in a way that's indistinguishable from reality and present it to these people.

and for those that aren't hyper aware, that becomes truth. And like you saw with the missiles or like we'll see with other campaigns, you can just create this world that appears real, has a lot of the same traits that would be real, but it's not. And that's a super dangerous thing. I'm not sure of any way to combat that. Yeah.

Yeah. This is a video I set you guys on Instagram. Oh, this one drove me crazy. So this is a video of a man jumping off of a cliff into a very small hole, which leads into water. And so he jumps into water, but the whole. hole that he jumps into is like six feet seven feet across and so he has to be incredibly precise he nails it right in the center but you just look at that and like all of us i don't think

any of us were like can definitively tell to tell that this is a i i or not we found the video of the creation of it and turns out he's jumping into a very big lake and then ai wrapped ground around the person and made it look incredibly real none of us were able to tell and so with the right tricks and if you're good enough at this, I think it's already possible to do even the most informed people about what is or what is not AI. So that reality is already here. David, there's another video as well, right, where like we have like some form of a host and she guides us through what's AI and what's not. Do you have that video? Okay, so here's the video that EJez was just talking about. So there are a number of examples of this lady talking about what's AI versus

not AI in her videos. And so it's kind of a quiz. So let's play this video and viewers, you have view to watch this. See how many you can get right. I got all but one right. Let's see how many you can get. In this video, see if you can tell what's real and what's AI. Thanks, Marissa. Okay. This bouncy house? It's AI, unfortunately. Looks fun though. Which landpost is real? That one's real. That one's AI. The bench, it's real. It's everyone's favorite game, Imperson potato. Which one's real? This one's real. This one's actually not a potato. It's a yam. There's no way you got that one right, David. I got that one right. They're real. I'll tell you on the left time at the end. No way. Is this gazebo real? No, it's actually this gazebo thing. What about this easel? You really got to try hard for these. The easel is real. I failed on 50% of these.

beautiful piece of art to hang in their home. You can't hang this because the art was AI generated. Sorry. Okay, so the way that I got that purse one is that you can tell the dimensions of the real purse change, whereas the other one, it looks like she's holding a static image. That's how you could tell. But other than that, like... Well, you need a trained eye.

You need a trained eye to know this. And I was using all of my brain to keep up with that one. Yeah. Yeah. This does appear to be the most dangerous part is the subtle nuance changes that you can't really disprove. Going back to the lake one, you sent it to me and like, I knew what was wrong intuitively. I was like, this is not right.

Intuitively, you knew it was wrong. But I couldn't prove it. And I think that was the part that is scary for me where like you have the Iranian missiles. You could kind of prove some of them, but it's getting close to the point where you can't. And if it's subtle adjustments that meaningfully change a story, it's really difficult to tell. And we're at that point like right now. Okay, so I've compiled like a pros and cons list, a constraints and superpowers list, of V-O-3 that I kind of want to just run by you guys.

So starting with the constraints. V-O-3, they can only produce clips that are eight seconds long. And so it's kind of a giveaway if we're trying to actually recreate real life. It's a giveaway if these clips are eight seconds long. All the subjects are more or less the same distance from the camera. And so they're within a particular bound of where the subject is from where the camera is. And so you can kind of tell that way. Continuity across scenes are pretty bad.

And so this is actually why gorilla or there's some bigfoot content out there, but also stormtroopers, like these subjects are very good because V-O-3, when it renders like a new scene, it re-rolls all the properties of the subject. And so if you are talking about like a human subject, you say like blonde curly-haired reporter, it's going to make a blonde curly-haired reporter that's new every single time. But if you just do a stormtrooper, the stormtrooper continuity continuity is strong. Otherwise, continuity cross scenes is pretty broken. And so if you're, you trying to like recreate the same scene twice. That's pretty difficult. Google does do a pretty good fix with this though, is that you can take the last scene from one clip, the last frame, and that will be the reference frame for the start frame of the next clip. And so they do solve that problem in a pretty clever way.

So you can't do like a cut scene with the same character as what you're saying. But if it was a continuous video, then you could. It doesn't, it doesn't look like a continuous video, but it is, it does kind of just like help the continuity of like the back. will be more or less the same. The subject will be closer to being the same. And so they have

a pretty good set of tricks to overcome this. And so those are those are kind of the constraints. Superpowers, 10x time, like 10x faster storyboard to render cycles. So you can just iterate so many times faster. If you don't like something, you can just do it again. Also, there's no need to license stock SFX or voice acting at all because VO just bakes in the audio. It creates the audio and dialogue in the model. So there's like, if we were to use a song, from somewhere, Spotify, in this video, we would get flagged for copyright infringement. Since V-O-3 just makes everything from scratch,

there's no copyright issues. So you're kind of like unlocked as a creator. Those are my big like kind of constraints and superpowers. Do you guys have anything to add to that? I was just thinking of another indirect pro, which is just, and Vida just keeps winning from all this stuff guys. Like I was just thinking about how much compute

is required to run or create one of these eight seconds clips and I'm guessing it's like quite large right now and it'll drive down costs over time. But like Jevon's paradox, which is the paradox where like, you know, if you create too much supply of one thing, eventually its cost or value goes down because there's so much of it, right? And not everyone wants to use it up. Actually, the opposite happens. And people find, yeah, people induce demand. They find different things to indulge in and basically it just, you basically end up with more people

wanting to use compute. NVIDIA is just benefiting from all of these different things. Whether we talk about V-O-3, whether we talk about a new model, whether we talk about Chad GBT's new image generation or whatever. It's just like, Nvidia is just absolutely crushing on this side. So it's an indirect little thing. All right, here my big takeaways as to what to watch for next. This is what I'm looking for downstream of V-O-3.

First, how quickly will society adapt to V-O-3 outputs? So we already saw Kalshi making the first ad. What if, like, people realize that these are V-O-3? output ads. They just intuitively know it just because of the style of the video content. And then they understands like, oh, this is AI. And because I know that it's AI, I actually think that this is a low quality product because they're not using humans. Like, what if humans adapt to the meta? Because it's so easy to produce these things that starts being a commodity. And like, you know, the third or fourth or fifth company to use V-O-3 to make an advertisement, sometimes maybe that starts

become like a negative signal. That's just one example of what could happen. How will society adapt to the fact that V-O-3 becomes a very commoditizable output. That's one thing. Tooling ecosystem, so how will tooling around V-O-3 and other AI video generators come to make these things higher quality, less sloppy, less discernible? And so just like tooling will emerge around these things, to just make these products better. Fundamentally, how good can V-O-3 get?

Can AI model working memory improve to be able to produce content that's just longer than eight seconds? Can the working memory of these models just do better, than eight seconds. Can we make 30 second clips that still retain continuity and doesn't look like manic? And then also creator anxiety, I'll call it creator anxiety. Mid-tier YouTubers will be undercut by anyone just willing to like churn and burn out content with Vio. So human creators who want to stay human and not just go down the AI generated rabbit hole, we'll need to double down on personal brand and just instinctive trust towards real humans. And then if you aren't willing to

make that gap, then you're going to go into the world of V-O-3 and you're going to just, like, figure out how to best leverage V-O-3 to make your YouTube channel or TikTok or Instagram Reels as good as possible. So those are kind of my what-to-watch things. Anything you guys are watching for? I mean, Alphabet that owns Google and YouTube, it seems obvious that they're going to integrate V-O-3 into YouTube at some point soon, right? To your point around tools, David. And then the other kind of thing I'm thinking about is, I think a lot of human YouTube. You mentioned, like, you know, they might just double down on human-made things. I think a lot of them will go over to the dark side.

Because, like, I'm just thinking about, like, how Instagram changed the game with filters, you know, and everyone just sort of, like, slapping filters on themselves. Snapchat did the same thing. I think we're going to see something very similar. Just from that clip that we showed earlier about that girl that was showing, you know, what was real and what's not. And one of the scenes was her sitting on a bench, which turned out to actually just be her friend, you know, just kind of like on all fours, yeah.

on the ground. I feel like people are going to get really creative with that. We're going to end up with some really weird human AI hybrid. Yeah. I'm personally excited for the, the pseudonymous type content. I think we see this a lot on Twitch, where some of the top Twitch streamers aren't real people, or they're real people disguised as these like virtual animated characters. Yeah. And when you remove the human from that loop and you're able to do that much more cheaply and much more effectively, I'm interested in seeing what that looks like because already we've started to condition people in the world of video games at least

to expect people that don't look like people but speak like people and move and act like people. So if you actually just replace the human behind it with this AI video-generated model, that probably doesn't change a whole lot in the eyes of the viewer, but in terms of the amount that you can create, that's a huge unlock.

So that's what I'm personally excited to see. It probably becomes more expressive, doesn't it? Like imagine that you have, this kind of mask to hide behind, like you're probably going to be kind of, I don't know, more honest or more transparent about like what your real thoughts or creative ideas are. And also, you know, yeah, more of the time. Like you can now stream 24 hours a day. There's no limitation. And you could also stream in different languages. You could take that version of yourself and create an Asian version that it happens into that market. That's really

large. And there's this whole unlock where you can just duplicate yourself. Oh, David's pulling up a black mirror trailer. This should be good. Do you guys know this episode, the Black Mirror episode, where this animated character who's controlled by a human and there's just technology that takes the human's emotional expressions and puts it onto this animated character, and then this animated character just

grows in popularity and then becomes elected president. Like, we have the technology to get here. We can do this. We can do this. We can do this now. Yeah. Yeah. Which actually, I think will leave me to the last big takeaway that I have, the last big thing that I'm watching,

which will be the future of memes. So memes are just a technology. We first had, like, you guys remember rage comics back in the early days of memes on the internet? They were just very reductive, very primitive forms of memes and, you know, just the first viral memes. You had the picture memes. You had, like, the awkward frog meme or, like,

all those OG memes that came well before Zoomers were even, like, aware of life. Then, you know, meme technology, progresses and now we have different meme formats because we like bandwidth on the internet got better phones and cameras got better and so the meme technology meme form factors got better and so I think the meme form factor is just going to get enhanced by Google V03 and there's going to be this like yeah just a new way of creating and sharing and going viral with memes when everyone will have like a video generator built into Instagram built into Google built into chat TPT and they'll be able to make memes

with video generation. And though I think like AI generated videos are going to be the new container form factor for memes that you're going to see like spreading around TikTok and Instagram real. So that's actually what I'm excited for. I'm excited for the memes. Guys, I'm pretty excited for our interview with PJ Ace.

He's the guy who created the Kalshi video, the Kalshi ad that went on the NBA finals. So I'm pretty excited for that interview. If you are just tuning into this channel, make sure to like and subscribe the video so you can go here from PJ Ace when we have him on. And if you are also interested in just getting this podcast on your RSS feed,

we are also wherever you get your podcast, Josh, Jaws. Thank you once again for going through another episode. Pleasure and awesome. See you guys soon.