The 8-Year Journey to Building a $4B AI Company | Synthesia, Victor Riparbelli

Intro

00:00So DIA is the world's leading AI video platform. We crossed 100 million AR. We launched the first product to a million AR. It took us 4 months. We then went from 1 million AR to 3 million AR in one year. But we basically tripled revenue from there.

00:11Tripled again and then we doubled and doubled. Are we doing this because it's fun and new? Are we doing this because it's actually useful for the customer? People are not trying to like use a model to do something. They're trying to solve a problem that they have.

00:21Right? In our case, the problem is making a video for something. I think a lot of AI apps are feeling this today where you may go viral, have a lot of people who are interested in your product and they come to you and play around with it. But it's really important to understand who are coming because they have a real use case that's recurring and who's coming because it's a cool fun thing that they want to try out.

00:39Do we actually think that they look real enough to work or is it a gimmick? Sometimes you'll be the hottest company in town and everybody will want to talk to you. Other times you'll be deemed as the incumbent that nobody cares about or like you're not doing something that's fun or kind of in the current sort of hype cycle.

00:53And no matter like which of those cycles you're in, I think the most important thing is really just like Victor Roelli and I'm one of the co-founders and the CEO of Synthesia. Sadesia is the world's leading AI video platform for the enterprise and for businesses.

01:09At its core, really what we help our customers do to land whatever message they have for their customers, their employees, their partners with the most engagement and effectiveness possible. So we've built a platform that enables anyone who knows how to use a PowerPoint to create a video.

01:22We're really what we're all about, right, is communicating in video, which is the most effective way to deliver a message. Today, we crossed 100 million AR. We've raised $330 million in venture capital. Commercially, we work with more than 70% of the Fortune 100.

01:35Today, we have more than 65,000 customers all around the world. People are not trying to like use a model to do something. They're trying to solve a problem that they have, right? In our case, the problem is making a video for something. We started the company in 2017.

01:49The

Why I Spent 8 Years Betting on a Crazy Idea

01:51first three years of the company, it wasn't very fun. and the tech didn't really work and it was it was a kind of an uphill battle. So this was like the first time the world really saw a neural network generate video at a pretty high quality fully automatically right if you saw the output of that face tof face video and you would have to do that in a visual effect studio you would be talking like millions of dollars to create a clip like this and this thing actually did it in real time and when I saw that for the first time my mind started running I felt like this going

02:14to change everything about how we create video but it's also going to change everything about video itself as a medium but I felt like it was worth exploring starting a company around this technology when we see new technologies the first thing we do always imagine all the 1,000 ways it can go wrong, right?

02:29Cars, smartphones, AI, video generation, music generation. That's always what we default to. But it just turns out most of the time that technologies end up being used mostly for good things by good people. And there's a subset of people use it for bad things.

02:40And I think being optimistic about technology, especially in the early stages of it, I think is actually a huge arbitrage in itself. I mean, essentially we went out right and we we told people look in 10 years you're going to be able to make a Hollywood film from your laptop just with your imagination.

02:54you won't need anything else than that and it's got to be more or less free to do it. And a lot of people just, you know, thought that there was a crazy idea. It was just it was in the AI winter. You know, there had been a lot of like disillusionment like a bunch of companies funding in the years before we started the company had all kind of like turned out to not really work because technology was just too immature.

03:13We got turned down by like 100 investors, something like that until we eventually sent a cold email to Mark Cuban. He was probably the only person we met who totally shared our vision of the future. So he wasn't evaluating the vision. He was bought in on that.

03:24He was more kind of evaluating if he thought the team was worth funding and if you think we had a shot at at building something real. You need one person to say yes. That's all it takes. And generally it's easier to try and find people who already share kind of like your vision of of what's going to happen in the world as opposed to trying to convince them of it.

03:40Especially if you're not doing something that's in the kind of current hype cycle. If in your gut you truly really feel that you're on to something and that someone if it's not you someone is going to build a big company around your the product that you're doing you should continue throughout the entire journey of synthesis I've never a day in my life doubted that someone is going to build a big company around this technology I could doubt myself sometimes I could doubt by execution I could be afraid that I was too early for where we were but I never doubted that this technology

04:08was going to be extremely valuable for the world and someone was going to build a company around

Are You Building a Painkiller or a Vitamin?

04:15The hard thing is knowing when to continue and when to pivot, right? We started off with a million dollars, right? Which is is not a lot of money when you're building a company like what we were trying to build. And so we had to figure out like what do we think we can realistically both build and sell before having to raise the next round, right?

04:30Which we would have to show some kind of progress. And the first thing we we we stumbled upon was this idea of AI dubbing, which is give us a real video shot with an actual camera. we can translate it for you by replacing the voice over with a different language but then also reanimating the face so it looks like uh the actor speaking a different language which will be useful if you're making a Hollywood film and you want to distribute that all over the world if you can make it look like it was recorded in French Italian that would be good for those audiences and

04:54that value proposition was actually good enough like that is definitely valuable when you can do it but the technology was very early it only worked if you look straight at the camera it took two PhDs you know like 10 days to make a 30-se secondond clip um it didn't scale at all the quality threshold was insanely high, right?

05:09Cuz all these video people, they wouldn't accept anything but like the very best in terms of of the quality and it was just clear to us that it was a vitamin. It wasn't a painkiller, right? Like people liked that. They thought it was cool. Um but if we disappeared, no one is going to yell and scream, you know?

05:23They are, you know, we'll go back to translating the way we did things before. So we went back to the drawing board and we of course talked to lots of people during this period. We kind of realized that rather than focusing on the people who are already making video all the time, there was a huge group of people, billions of people probably that weren't making videos at all.

05:39And these people are desperate to make video. And they were basically telling us, hey, I want to deliver this message to someone. And I know I should be doing a video or audio because that's what people want to consume. I just don't know where to start.

05:50I don't know how to use a camera. My boss won't give me any budget to record it. And I know that there's so many problems after I've actually shot the content even though I get to make a video because you can't edit a video after you've shot it.

06:00And there's all these problems of of video, right? And then we we spoke more and more to these people and ultimately learned a lot what these people wanted to do was like corporate videos. And corporate videos are a lot easier than Steven Spielberg filmed, right?

06:12And so if we could just essentially help our customers with kind of like the talking head style footage, that would be pretty big. And that was a much smaller domain than make any kind of Hollywood film you can imagine, right? So we built this avatar technology that became one of the central of the product today.

06:24That's when the company started to really take off very quickly. We had a lot of uh conviction on a technology very early. It was very unclear what the right first use case was. You know, something that was both valuable and you could realistically build.

06:38But I think we found our way there. You could argue that we should just have built the avatar technology from day one. But back that day one, none of us had any idea how we would potentially do that, right? the research evolved over the years.

06:51We learned the technology better and better and so I think product market fit is one of those kind of things which is is much more art than science and I think we're very much led by by deep understanding of the customers that we're serving and of course deep insight into into the technologies.

07:06It it ultimately turned out the right ways.

Fun and Cool Won't Keep Users Coming Back

07:12Once we launched the product and we had the first avatar MVP out there, we went viral very quickly. And so we had like lots of people coming onto the website, lots of people making kind of free demo videos of avatars by typing in things because it was fun and cool and they want to show all their friends and it's amazing.

07:27The first time you see it, you're kind of mind blown, right? But a lot of these people did not have any real use case and we could see that in the retention, right? They weren't coming back. They were just making a few videos, showing it to people, not making any more videos.

07:37And so that wasn't really like a sustainable revenue stream for us. But there was a small group of people and they kept using the product you know they kept coming back and uh we we we kind of went into that group and that spent a lot of time speaking to those folks and that's where we uncovered sort of interesting value equation right where these people were not comparing our videos to real videos or camera videos they're comparing our videos to text documents and when that's the comparison then the quality threshold looks very different right the feature request that they needed look

08:05very different so it was an organic thing I think of of letting yourself be guided by customers but also applying a lot of like critical thinking. The critical thinking is like difficult when you have so much inbound interest. I think a lot of AI apps are feeling this today where you may go viral have a lot of people who are interested in your product and that come to you and play around with it.

08:21But it's really important to understand who are coming because they have a real use case that's recurring and who's coming because it's a cool fun thing that they want to try out. Early days of 2021, we have customers who will scream and yell if we shut down the product tomorrow.

08:34That's always I think a good asset test of like do you have something real or not or do people kind of mildly care but not really. We all realize that there are lots of unanswered questions for sure but there clearly is something right there's a kernel of something that is uh that is huge and I think it was kind of double exciting in some ways because it felt like it was our little secret and it was for a couple of years it wasn't very obvious from the outside that anyone had any real use for this tech.

08:58That was a fun couple of years where we felt like we know something the rest of the world doesn't and and we're capitalizing on it.

The Vacuum Effect of Product-Led Growth

09:05We figured very early on actually in 2021 that we wanted to focus in the enterprise because that was clearly where there was the most value to drive and that was clearly where we could you know get to the kind of biggest contract size. But it was also very obvious to us that the growth channels and the way to get this in the hands as many people as possible was not through enterprise sales.

09:20It was through just giving people access to the product, letting them play around with it, let them find their own use case, letting them qualify themselves before they spoke to the sales team. What was hard is when you're building both bottoms up motion and you're trying to build a tops down motion.

09:33Those things conflict in many different ways. As a startup, you have limited resource. And a lot of things you build for a big company will have very little impact on a small company, right? Like if you build admin tooling uh for like a CIO, a 50% company is not going to care about that.

09:47So they won't get much value out of it. On the other hand, if you build tools for like small businesses, a lot of that will not have value for someone who is um is really big business, right? There's always like these trade-offs that you make.

09:56But I think when you look at the last, you know, 10-15 years, some of the companies that have had like the best and most impressive runs are companies that have managed to combine the kind of vacuum effect of PLG where when you really get it working, right, it's just this vortex that just like sucks in traffic and convert it into both paying customers and credit cards, but also spits out, you know, the world's best enterprise leads with 1010 worldclass sales team that knows how to handhel customers, take them through a sales process, make them successful, and do all these different things.

10:25And it's interesting because it's not just in the expression of the product and the product roadmap. A lot of people tend to think that, but it's also very much in just like the people you hire, right? Some people really love like the self-service world and that's what they're passionate about.

10:37Other people really love enterprise sales. And so marrying like those two cultures together is definitely a challenge, but um I think we've in broadstrokes done pretty well on it. I think a mistake people often make

Find the Bit-Off Underdogs

10:50especially people who maybe have already had a job in like big tech or like a successful companies they go out and then they try to hire people from Google, Meta, OpenAI like you know whatever kind of is the hot companies of our time and they go out they hey I have this cool startup idea why don't you join me I'll pay you not as much but I'll try and pay you almost roughly as much etc.

11:08And of course this can work. There are very smart people who work there who also wants to the startup hustle. But often times you just can't get them straight up or if you do get them and they are used to a very different salary, a very different lifestyle uh working in a big tech company as opposed to being in a startup.

11:22So that's actually often not the best path. So I think that the better way of doing it, which is what we did, is you try and find people that are a bit off. You know, they're not like the most obvious people to join your startup, but there's something you like about them.

11:34That could be that they have just a lot of hustle, a lot of grit engineers that maybe haven't worked at like a tier one or tier two company, but they are great open source project that they're managing and updating. Like there are a lot of other signals you can find good people, right?

11:46And if you can find these people who are truly hungry and wants to work at a startup and wants to put in the work, you gather a great group of those and they will very often outperform the 10 people who used to work in big tech uh with their comfy lifestyles.

12:01people you can take and we help shape give them opportunities to do something that's your only option in the early days finding that like racktac group of underdogs that have the collective energy to to go up against the competitors and incumbents I think that there is nothing that prepares you to really start a business like starting a business like spending 4 years in Mckenzie is not going to make you a better entrepreneur make you have a little bit more network and maybe slightly easy to raise money but if you actually truly if you know deep in your heart that you want to build a company

12:28and you are a builder I think you just get started as quickly as you And don't be afraid of starting too early. The the doom scenario is is rarely as bad as people kind of think it is or make it out to be. And most big companies really value people who have tried to do something themselves even if they failed.

12:43I do think that we will see less and less text in the world. I don't think we will be consuming most of our information in text. I think there'll be still be a place for text. You know, we still go to the theater, we still listen to vinyl. Like there'll there'll definitely be a place for text to exist.

12:56But I think in in the world of information sharing, training knowledge, etc. I do think that we will move into a world which we consume almost everything as video and audio and maybe VR and AR um in in 10 20 years with those technologies. I do think that's going to happen like and I don't think it's going to be as bad as people think it is.

13:10A lot of people are very like averse to the idea that people wouldn't necessarily be using tech as much as as they are today. But I think as humans it's very clear that we always been on this trajectory towards more interactive and more visual and more auditory content that reminds us more of like actually being in the real world, right?