TRANSKRYPCJA 🔗
Victor Riparbelli (00:00):
So I'm Victor Riparbellii, and I am the CEO and one of the co-founders of Synthesia. At our core for almost the last 10 years, we've helped people make video without having to use cameras, studios, microphones, actors to generate everything from behind your laptop. We built, you know, the world's biggest platform for the enterprise to create AI videos. We work with 95% of Fortune 100. And then we just launched our new products, which is real-time video, AI native video that we're very excited about, which is an evolution of the video format, moving it away from just being a broadcast that you passively consume, to a format you can actually interact with. So if you're training someone on something, rather than than just learning by watching a video, they can now have a conversation with an avatar who can act like an examinator or tutor or whatever you want it to be.
Victor Riparbelli (00:48):
So we started a company in Europe. We are headquartered in London and we have offices in, you know, quite a lot of countries around Europe by now, the small offices. And then we have New York and we have Austin. I do think Europe is, is, is somewhat moving in the right direction in the AI, AI era. What we have seen is that you can build global leaders from Europe. So you look at companies like ElevenLabs, uh, Wave are popular in there, companies like us. We're not just like the European player for something, but actually is winning on a global stage. So that's very positive. Lovable is another good example of that. But unfortunately, it is still just like a incredibly small percentage of the overall value being generated in AI, right? Like obviously the US is way ahead. China is the very clear second competitor. Um, and I think there's a, there's a lot of great people in Europe, a lot of great talent. I don't think that the EU understands what it's gonna take to win. And I think there's too many people who live in this kind of bubble. You know, in the real world, it is a competition. We need to arm our companies to win. It's not a luxury to have a great economy. It's a necessity if you wanna keep the standard of life that we have. And we're starting to see some cracks in that. But I think in 10 or 20 years, it'll be very obvious that Europe got left behind in not just the first tech revolution, but also the second tech revolution. And that's when we're gonna start to see like real problems across Europe. And I think it's, it's even very simple things like a lot of the regulation, the governance pieces, right? Like the most powerful models are available in Europe often two or three months after they're available in the US because the companies from the US have to comply with all these like checkbox exercises. And I understand that there's, uh, and I can emphasize to some extent with why some of they want, they want some of this governance. I think in reality, I don't think it matters. I think it's just like bureaucratic paper shuffling back and forth. And it's a massive hindrance, right? Like two months in AI right now of not having access to the best model is five years in the previous generation of, of software. I think all of this stuff cascades down from culture and I love European culture. I feel very European myself, but I don't know what we really believe in in Europe anymore. You know, in, in, in the US they believe in capitalism for good and for bad. In China, they believe in technology, like they really love technology. I'm not sure what believing in Europe anymore. Maybe it's regulation, maybe it's climate catastrophe, maybe it's AI dooming, I don't know. But there, those are not beliefs that drives new entrepreneurs, that drives like healthy markets where you can build a business and you can grow really fast. And I think that's a, a big problem. You know, I've, I've before described Europe as the trust fund kit of the world, you know, we inherited a lot of wealth. I think we have some level of back conscious of like some of the stuff that sort of went down history over the last hundred years. And, uh, because we have great lifestyles and great economies, we then all become like very idealistic about the world and we forgot that unfortunately in the real world, like we need to have great economies and great economies comes with some trade offs. But there's also much smaller companies, right? Like if you look at market cap of like European companies, if you speak to most VCs, they, they won't say it in, in, in public, right? But right now, maybe there's like one or two companies that's interesting in Europe every year as like they have real potential to become actual multi-billion dollar global leaders, and that's what VCs are, are vying for, right? If you look in San Francisco, it's a lot. And it is also much harder. I, I think, you know, I'm very proud of, of everything we've done, um, and we wanna stay in Europe, but there are real challenges. Like there's of course all the sort of structural challenges, but there's even just, uh, because we don't have a very mature ecosystem, once you actually get to some level of scale, it's very hard to find, you know, senior talent that have seen how to take a company from hundreds of millions of ARs to billions of dollars of AR. And those are the things we... Like what, what Europe needs is, of course we need a higher quantity of startups to some extent, but what we really need is like huge, massive companies, right, that really, truly drive the ecosystem. Uh, like we even most Spotifys, right? We only have one Spotify in Europe. And if you look at just the impact that has had in Sweden as a startup nation, it's incredible, you know, Legora, Klana, uh, Lovable, like there's so many great companies. And a big part of that I think is because Spotify made it really big, you know, people aspire to like build something similar. If you go to my home country, Denmark, there's nothing. You cannot name me a billion dollar
Load full transcript
Danish tech startup. I think that is like the ecosystem part. I think because we, because we haven't had a really big success in Denmark and because Denmark is notoriously difficult to build a company in, there's no ecosystem. I have some optimism. I mean, if you caught me a year ago, a one and a half year ago, I was a lot more optimistic because I felt like there was more like, okay, we screwed up. Let's do this now. And then I think what's happened since then is just a lot of Talk and more regulation and more bad decisions. Um, you know, it's like Europe tries to fund these like sovereign labs for LLMs, right? And I understand like the sort of idea behind it. The problem is that in an ideal world, yeah, maybe sure, we can go and train our own LM models that we then own, right? But the first problem is that they're gonna be hosted in American clouds that is still like the foundation of the ecosystem. The second thing is just that if you're running a business, right, it's the worst possible strategy to try and compete with LLMs because all businesses will pick the best LLM model at the cheapest cost. Like a lot of companies will like posture as saying they wanna use European tech. We haven't seen that yet. And I don't think when you're being out competed by China and the US that you're just gonna, out of the goodness of your heart, gonna choose a European provider. It's just unrealistic, right? And so the government pumps like money into like supercomputers. Um, we have nowhere in the capital that's available in the US. We have nowhere near the talent available to do it. And so what's gonna happen is we're gonna build probably an LLM that's gonna be the like 15th best in the world and nobody's gonna use it and we're gonna burn billions of dollars on it. Instead of just saying, "Let's focus on the companies here that are actually working and make them really big." I think what Europe has, right? I think we have a lot of like culture, a lot of creativity, and I think you can act to some extent that a lot of the, I think next generation of products that we're gonna be building, I think especially with things like, you know, AI generate art and video and this sort of stuff, I think Europe actually has an edge there. Taste is a... And, and you know, a lot of people talk about like what's gonna be ultimately very important age of AI, it's going to be taste. I think we can make it a bright future. I just really hope that the EU kind of pulls themselves together and start putting in roadblocks and start to like see the optimism. I think people love like the startup in the garage because it's kind of a revolved against like the big evil capitalism system. But then as soon as these companies actually start working and they become like big companies, then all of a sudden people are like, uh, now we don't like them anymore because now it's actually like a big corporation, right?We started the company 10 years ago, and we started it based off one of my co-founders, very, very early AI video research back in, in 2016. He did a paper called Face-to-Face. And when I saw that paper, I was like, "If this progress continues, two things are gonna happen. The first one is gonna be the obvious one. You're gonna be able to produce, you know, video content from behind your laptop, just like you could, at that point, almost create any image if you're good enough in Photoshop, you could create any song you want by using virtual instruments and just your MacBook. Ultimately, you can write a book by just typing on your keyboard. That's gonna happen to, to video as well. And I think we're very square in that right now. There's still a lot of progress to do. And then the second big idea was the idea that what always happens when we invent new media technology is that the media formats themselves change, right?
Victor Riparbelli (01:32):
Like when we had the advent of computers and the internet, we didn't just all start like writing books and put them on websites. The idea of a block was invented. And now we're at like TikTok videos and podcasts and all these new formats. And AI will enable a whole bunch of those. Um, so those are the two big ideas back then. Now, a lot has happened since we started the company. The first three or four years, we were kind of in sort of a research lab mode where we're just trying to figure out what could we potentially build that would have some value, even though the technology is like barely worked. And we had our first breakthrough on that in the end of 2020. We invented this avatar technology, which basically re- reduced the need for a camera. You could just type out a script and you would have a person sort of reading that out.
Victor Riparbelli (02:13):
We then built it into a video editor. We enabled essentially the world's office users to think like your average PowerPoint, uh, user in a big company to create video, essentially for them to circumvent having to work with an agency or with video experts. Now the quality of those videos were much lower. It felt more like a PowerPoint in a video format. It wasn't like a super well edited video with motion graphics and cinematic shots and all those sort of stuff. And what's happened since then, especially the last 12 months, is that the video models have gotten to such a point now that I think we're almost past the point of technical limitations of creating something that looks like a Hollywood film and you barely can tell. And if we aren't there yet, we're going to be there in six to 12 months. The core capability of like the type of video we can create has increased massively in terms of the cinematic aspect of it. Motion graphics
Victor Riparbelli (03:00):
And all the... I, I usually explain it as like if you take a, let's say a YouTube video of, from a great content creator, what is that video? It is ultimately a bunch of assets. Maybe something was recorded on a camera, maybe there's some animations, right? It's an
animated bar chart, you see text that moves around. Like all the things is that makes a video a really good video. Up until now we've kind of been able to take the individual assets and create those with AI, but you still would need someone who's a really good video editor to sit down and select the right soundtrack that matches the assets, do the motion graphics, write the script, all those sort of things of putting the individual blocks of a video together was really hard. There's been a massive breakthrough. We just launched our product called Siren where basically what we're doing is we're using an agent that kind of works a bit like a human video editor.
Victor Riparbelli (03:46):
You give it a task, you explain what video you want to create, and it can create like literally any type of video you want to create. If you want a cinematic Hollywood looking thing or if you want like purely just motion graphics thing, you just explain what you want to create. You can give it documents or web links and so on. And what it'll actually do then is it'll go out and it'll have access to all these different models it can use. It has access to LLMs, it can write a script, it has access to our avatar models if you want to have, you know, person talking to the camera, access to third party video models if you want something that our internal video models, you know, cannot create. And then what it does, it actually works like a human video. It puts together a video with animations, Music, sound effects, and the quality of those videos, like absolutely crazy how good they are. And then the way that you kind of work with it is like you would with a human video editor, you chat with it. So you can say, "Hey, like around 30 seconds, change the text here or make the section a little bit shorter, make this a little bit longer, change, you know, the avatar to be slightly older, younger, whatever it is that you want it to, to be." And so what it enables is that any person on the planet now can create amazing looking video simply by chatting, which is a very, very natural form of creating from a human perspective because it's kind of like working with a colleague. Yeah. And if you look at like the, the kind of consumption side of, of things, right? Like if you go on any social media platform right now, there's a ton of AI content and I think, you know, it ranges from the very, very AI slop, crap, low quality, spammy type of content to the most amazing, surreal AI art done by artists captivating like millions of people, which mirrors like all the other content you find on internet, but there's a lot of crap and there's a lot of like really good stuff, right?
Victor Riparbelli (05:23):
But it's got become part of the mainstream now. And I do think as much as the video space has grown very significantly over the last 12 months, I think this is sort of the ChatGPT moment for video. And I think that over the coming years, I think it's gotta be hard to fathom how much video is gonna be created and consumed and how big of an industry is going to be.
Victor Riparbelli (06:00):
10 years ago, the tagline for the company was, "In 10 years you can make a Hollywood film from your laptop with just your imagination." That's kind of what I wanted to do because I, I love creative things, you know, but we were very aware that at that point in time that was like definitely not going to be possible. But we started off in trying to create tools for like Hollywood studios, video professionals to not create full AI things, but just kind of assist them. We, we learned after a couple of years that that was a bad path for the company because all these folks who are into great video creators, very skeptical of any new tools, very skeptical of AI, super high quality thresholds to do anything. And so what we saw was that there was a lot of interest from non-video creators. A lot of those sat in businesses where they were definitely not big in Hollywood films, they were making like onboarding training videos for new people who joined their team.
Victor Riparbelli (06:50):
Or they needed to create an academy of, for their customers around like how we use their product. They're all telling us the same thing. It's like, "Hey, I know I need to create video. People wanna watch and listen. My long doctrine is not being read, sending a 25-page slide deck. My message gets lost. People wanna watch and listen, especially of course the younger generation." And so we saw an opening there of like all these people, if we can enable them to create video, that'd be great. And we probably can because their needs are very different. You know, this is like high volume of video at a much lower quality, you know? And the way we though about the quality threshold, and I, I think one of the really big early insights was this thing that if you speak to professional video people, the quality buy they'll hold it against is an actual real video because that's what they can do.
Victor Riparbelli (07:34):
That's their alternative, right? But for all these people who are PowerPoint users essentially, their alternative was not a video. It was a document. And so even if the AI videos were not of very high quality, as long as they were better than the 10-page document, it was a win. And so we built a whole company around that. Um, and to this day, I think, you know, we still are... We're not gonna be the company, I think, that professional creatives who's creating like your Super Bowl ad or Hollywood film, something like that, we're not gonna be the platform for that. We are gonna be the platform that enables like anyone to create a video really quickly and really fast. And I think if you look at the history of technology, right, it's like, it's just, it's a very obvious arc. There was at some point in time, like actually writing things was not part of an average person's job, right?
Victor Riparbelli (08:23):
You would have like secretaries, you would have typewriters, and it was their job to sit down and actually create text documents. Then we all had computers and laptops. Then we all kind of slowly became designers with the advent of PowerPoint, right? Like most office workers today may not be like the world's best PowerPoint designer, but everyone can like put together a PowerPoint. And that is actually design and visual communication, even though we don't think of it like that ourselves. And now what's gonna happen is we're gonna be able to create videos. We're already seeing how people are able to create agents with LLMs, which is sort of like software developing. So we're just seeing this crazy flattening
Victor Riparbelli (08:54):
Of like a lot of the very, very specialist skills that existed 10, 15 years ago, um, are much, much more accessible than they are today.