Victor Riparbelli, the Danish chief executive and founder of Synthesia, is on a mission to reinvent video content.
Synthesia, which was set up in London in 2017 as a platform that used AI to offer lip-syncing and dubbing services for film and TV productions, has grown to become one of the UK’s most valuable AI start-ups. A funding round in late January valued it at $4bn.
The company has developed tools that lets users upload photos of themselves in order to create avatars that can render training videos in multiple languages.
In conversation with the Financial Times’ AI correspondent Melissa Heikkilä, Riparbelli says that the way people interact with video content will change radically, just as the costs fall. He also talks more broadly about how Europe can compete in the global AI race.
Melissa Heikkilä: Tell me, what is Synthesia, and what do you guys do?
Victor Riparbelli: Synthesia is the world’s largest AI video platform for businesses. And really what we do is we help our customers communicate faster, better, and with more engagement than they could before. That’s, of course, all about video. Five years ago we were the first to take a new AI video technology to market that, essentially, enables our customers to create videos without having to use cameras, studios, microphones, and all this physical production process, which, to most people, is how we think of how we make video.
Instead, we can use these AI models to actually generate the pixels. You can create an avatar of yourself. You can use one of our stock avatars available on the platform. And if you’re creating these talking-head-style videos, you just type in the script, and then we give you a video of that. Since then, the platform has evolved a lot. The AI models have gotten better. You can generate anything you can imagine by just prompting. You have a full editing platform.
MH: And you’ve just recently raised a new round with a new $4bn valuation. Can you tell me a bit about where you’re heading next?
VR: We had two big ideas when we founded the company. The first one, which is playing out right now, and I would say that’s the business we’ve built today, is just you can create faster, better video by not using cameras, but by doing it with AI.
The cost of producing a piece of content is going to drop to zero, not just in the cost required, but also in the time and skill required to do something. And that’s what we’re seeing right now, play out, not just within Synthesia. [This is happening in] the whole AI video ecosystem, and it’s a huge business opportunity, and that in itself is going to transform the world.
The second big idea was that what always happens when we invent these new media technologies is that there is this bridge period, which I would say we’re still in now, where we invent this new technology, and then we use it to create media formats as they’ve always been.
So today, AI video is video as you know it, it’s a broadcast format. You create one video, you upload it to YouTube, and everybody watches exactly the same version of that video. It’s just made with AI, but there’s no difference in the video as a format.
That’s going to change, because the $100bn question in the space is what does video look like when it’s truly AI-native? Forget everything we know about video. We have LLMs [large language models], we have AI video models that can produce pixels, we have voice models and a whole bunch of other cool technologies. If we put all these together . . . and we rethink the experience of what is it like to teach someone something, what does that experience look like in an ideal world, that is going to look very different from the video that we know today.
What this means is that video is going to be personalised. Everybody’s not going to watch exactly the same version of a video. It’s going to be interactive.
So you can, in the case of what we’re doing with a lot of our training videos, rather than just passively consuming something and learning the content, go into a section of the video that has an AI tutor that you can ask a question. Or maybe the AI tutor role plays with you to make sure you’ve actually understood the content, if you’re a salesperson, for example. So that’s a very different learning experience than just watching a video.
MH: What sort of AI breakthroughs enabled this new technology? Because when transformers, the architecture that power LLMs, came out, that allowed you to do the avatars and the language tools you have now, but what’s happening now in AI that allows you to take this to the next level?
VR: There’s a bunch of different things. With these new videos, these are intelligent videos. Like an LLM, they have a brain, and they can talk to you, they can interview you, they can do a lot of things with you. So one of the core techniques that power this, of course, is the LLMs. So the rise in the quality of LLMs, the capability of LLMs, is a very important part of building this product, because it requires intelligence. And that’s not models we build ourselves, we work with big providers on this. So that’s one big unlock, just the models getting better.
For the video side, it’s things like having real-time avatars, for example. So rather than generating a video, which takes some minutes, instead what we can do is we can actually generate it in real time.
But it’s also about rethinking not just the avatar in real time, but let[ting] the video pull up relevant information, or draw[ing] information on the screen that’s relevant to what you’re talking about. If you’re teaching someone about a new pricing policy as a software salesperson, for example, what’s the most effective way of learning that? That’s probably that you draw a table showing how you price the product in different categories or size of business or whatever.
It is almost like these videos are becoming small apps. It’s just a lot of these technologies becoming better and better.
MH: One big question is it’s increasingly becoming harder to know what’s real and what’s not. How do I know, how do our viewers know, you’re real?
VR: I’m not. [Laughs.] That’s always been a huge discussion point since we founded the company back in 2017. The first five years, I couldn’t do a single interview without that being 80 per cent of what we talked about. This is going to be a problem. It is, to some extent, a problem today. It’s going to be a problem in different ways than most people imagine is going to be, because technologies have these weird ways of developing that are very hard to predict.
There’s recently been some cases, and most of the very serious cases that we’ve seen is around things like deepfakes for pornographic content, non-consensual pornography, those kind of things. That’s going to be a big problem.
For us, we do a lot of work on ensuring our technology isn’t misused. We do consent checks, so you can’t create avatars for anyone but yourself. We do content moderation, which means we take a stance on what content you can or cannot generate, and a bunch of other things.
In my own personal, anecdotal, experience, young people are really good at spotting what’s real and what’s not real. Older people have a very difficult time spotting what’s just AI slop videos and what’s actually real. That’s because the younger generation is much more exposed to these things, and they have it much closer to their daily life.
MH: Creative industries are very worried about AI taking over their jobs, how do you see this? Is this technology something that’s going to replace them or something else?
VR: The rhetoric around these things is always very extreme, at either end. The people in Silicon Valley [would say] yes, you’re never going to need anything else than just a prompt, a chat box, to do anything you want. And then, there’s other people who say nothing’s going to change at all, nobody will want to watch AI content.
And it’s got to be somewhere in the middle. If you look at the history of technology, jobs will definitely change. The way you create video will be very different. But I also don’t think all video is going to be AI video. If doing news reporting from an event around the world, that’s not going to be AI-generated.
What I see . . . is that using these systems to create content, three or four years ago, a lot of artists were very really disliking AI. ‘It’s bad, it’s going to have all this slop, it’s going to take our income stream,’ and so on and so forth. I have a lot of friends who are artists, also, in my own spare time, I make music, for example, so I’m experiencing it first-hand myself.
What is very true with these things is that the people who are by far the best at using them are the artists, the people who used to be, or are still, great photographers who know how to use Photoshop. Because a lot of this, what we see in Synthesia is it’s very hard to underestimate how little creativity most people have. It’s very, very hard for most people to come up with ideas and to evaluate those ideas and to know what great looks like.
What we’ll see is just that those people will become much, much, much more effective. They can storyboard much faster, they can do their job a lot faster. And just in my own, again, anecdotal circle of friends who are illustrators, photographers, I see a lot of them really picking it up and just making [it] a really big part of their workflow. Yes, there’s parts of it that they don’t like, but it speeds up their workflow so much, and they can do so much more with very little budget.
MH: You’re Danish, why did you want to start this AI company in the UK and stay in Europe, when I’m sure you had lots of offers in the US?
VR: I knew from an early age that I wanted to go out and explore the world. I love Denmark, I’m very proud to be from there. Denmark is a small country, [and] it’s a country that, culturally, doesn’t really celebrate drive, ambition, innovation. It’s like the polar opposite of the US, culturally, in terms of what does the culture care about, what’s celebrated, what is viewed as virtue, etc. I always just felt like I was a bit out of place in Denmark.
And when I went to the US, I did a semester at university, just experienced a very, very different ecosystem of people where I felt like people dare to dream big. It was very infectious . . . and it gave me a lot of inspiration. So part of it was because, on a personal level, I want[ed] to go out and explore the world. I couldn’t get to the US, because of visa issues, so I ended up here in London.
MH: You are building one of the biggest AI companies in the UK. How do you do that? The whole narrative in Europe is that we’re falling behind the US, and it’s really hard, there’s not enough money.
VR: We’ve over-regulated. We focused way too much on preventing imaginary harms than building strong economies with strong companies.
It’s gotten better, and it’s gotten better in part because we’ve had some tumultuous years, we’ve had wars, we’ve had geopolitical tensions with the US, and so it does feel like people are starting to realise that having a great economy is not a nice-to-have. It’s a need-to-have if you want to maintain the living standard that we have today.
And that seems to be seeping in a little bit more. You’re seeing that across the board from politicians being much more pro-business, pro-innovation, wanting to build local heroes. I actually feel like the government in the UK are not adversaries. That’s the feeling most entrepreneurs would have had up until a couple of years ago. They actually want to help. They’re proud of the companies that have been built here.
There’s a lot of great talent in Europe. If you’re a top-tier company in Europe, it’s a massive competitive advantage that you can hoover up all the great talent here, because there’s less competition. In the Valley, you’re competing with 600 other well-funded, cool start-ups doing cutting-edge things. In Europe, we don’t have as much, so you have a less competitive situation for talent.
But where it does become more challenging is when you need to hire senior folks, executives. We’re operating a company now where it’s not getting to how do we make $5mn ARR [annual recurring revenue] on a software product. It’s how do we get to $1bn of ARR? It’s a very different challenge. And you need people who have seen that journey play out before.
And we don’t have a lot of those in Europe, which means you have to go to the US. You have to try and find people there who are willing to move. And so, there [are] a lot of complications.
From a funding perspective, there’s definitely still a discount on European companies compared to American peers. It’s evened out to some extent. The capital markets [are] much more global now than they were in 2017. VCs, are, by and large, investing, anywhere in the western world. It’s important to me to build a big company in Europe. Net-net would it have been bigger, more successful if we were funded in the US? I don’t know, but structurally, it probably is still a better place to start a company.
MH: Are you seeing, with the political turmoil in the US, more interest in your company now? Are people moving?
VR: When you get these polarisations, you’ll have a group of people that’ll be very interested in staying, and they’ll be another group of people that are maybe more interested in moving out. We are seeing that as well, yes.
MH: Now, the UK is obviously very keen to be competitive in the AI space, what do you think the government needs to do to actually achieve that?
VR: The most important thing is let the markets be good markets to operate in, and not get too much in the way. Things like regulation, for example, the fact that we’re not in the EU is actually a very positive thing for the UK. The EU is now backtracking with the AI Act, and the latest draft I saw, I still think there’s a bunch of red flags in there. But in the UK, of course, if we operate in the EU, we still have to adhere to those guidelines, but as a place [to build from], that’s a good start.
There should be careful regulation. I’m pro regulation. We need something at some point, but I don’t think we should rush it. And as much as you like it or not, it is a big part of building competitive advantage. So that’s the first thing.
Then, we should be better at having governments, institutions use AI, especially from British providers. Show, don’t tell, that AI is going to be important. It will drive so much productivity and efficiency in the government, which is also very much needed.
Then, there’s all the wealth tax, how attractive is the UK to move to to fund a company? I personally don’t want to live my life based on where the tax is the lowest, but that is a factor for a lot of people.
And looking at the wealth tax, for example, you should discriminate between different types of wealth. The wealth that I would think most people in the country want is that people start companies, they start businesses, they create jobs, they bring in investments. And then, there’s another type of wealth which is more rent-seeking by nature, old families, old companies, just not being very productive to society, but still making a lot of money. Maybe it makes more sense to tax that.
[The government] removed the non-dom scheme, which was a very attractive thing for people to move to London. And having a good environment for people to build business in, that’s not just baseline competitive with other alternatives, but actually is better, is important.
And then, last but not least, there are some AI-specific opportunities. There are big public data sets in the NHS, for example, which you could utilise in a medical context. And there are other data sets that are probably very useful. If we sat down and we figured out those things, and a way to deploy into the NHS or school systems, or whatever, these AI systems, that would also be very beneficial.
The one thing to note here is that one thing I have not said is more data centres, which seems to be the number one thing that everybody is talking about: ‘We need more data centres, we need more data centres.’ Of course, we need data centres, but it’s a red herring. I don’t know a single AI company here that’s held back by the fact that there are no UK data centres. You’ll find your compute somewhere else.
We need to build out the compute, for sure, but I don’t think that’s a hindrance, especially because when you look at the ecosystem in the UK and in Europe specifically, we’re not talking about truly foundational model companies at the scale of OpenAI and Anthropic and those kind of things. Europe is very application data-centric, just like we are.
Yes, we use a lot of compute, but it’s not a bottleneck in the same way as it is for some of these big companies. But data centres is just a very easy thing to talk about, and it’s big numbers and it sounds good, but that’s the wrong focus.
MH: So then, why are people pushing this narrative, pushing data centres? Is it just the three companies that need it?
VR: People look at what the big tech companies in the US are saying is important. They’re looking at all the announcements that are being made, and they’re concluding that we have to do the same thing. And then, of course, there’s a whole industry of people who build data centres, who’ve had some very, very, very good years and are very, very interested in pushing this narrative that this the most important thing.
Again, to be clear, we need the infrastructure. Inference [a term to describe the process of AI models generating outputs] is, of course, going to scale massively in the next couple of years, but if you look at it from an ecosystem perspective, of how do we get more companies to be founded in the UK and be built in the UK, that has nothing to do with data centres.
MH: What’s next for Synthesia? And where are you hoping to take the technology next?
VR: The big focus for us, really, is moving from this video-as-we-know-it-but-made-with-AI to a new type of video that’s only possible with the use of AI and being AI-native, so these videos you can talk to.
We’re starting with our first product, which we’re calling Skills, that’s launching pretty soon, which is all about upskilling and training your staff.
One of the things I hear in every boardroom I’m in is that people are excited about all the new, cool things you can do with AI, but there’s also a lot of fear around how do I transition to this new economy. And a lot of people, executives look at their teams, look at their workforce, I have great people, but they don’t know how to use AI. How do I really quickly get them up to speed with how to operate in this new economy?
And so, everyone is thinking about how they win in this new economy. And I’m very bullish on people, actually. There are a lot of companies that talk a lot about automation, and there’s going to be no jobs left, everything’s just automation. There’s got to be so much automation, of course, but I still very, very much believe that the core DNA of great companies in the future is still going to be really great people.
The jobs will probably change, there’ll be a lot of automation, but the people [are] still what will make up a company. If you look at a lot of the companies pushing this narrative that it’s just all about automating everyone away, they’re the companies hiring the most themselves.
MH: Well, that’s a hopeful note for humans. Thank you so much, Victor.
This transcript has been edited for brevity and clarity