Synthesia reaches 90% of Fortune 100 with enterprise AI video platform that replaces text, not Hollywood

First AI company we've ever spoken to. Yeah, it is the first I've actually seen these uh these ads. Uh the Synthesia is the world's largest AI video platform for the enterprise. And so we'll see how this is all taking shape. Welcome to the show, Victor. How are you doing? What's going on? I'm very good, guys.

How are you? I'm great. Uh why don't you kick us off with an introduction on yourself, the business, kind of uh give us a lay of the land as it stands now? Sure. Yeah. I'm Victor. I'm one of the co-founders and and CEO of Cesia.

Um started the company in 2017, so it's kind of a while back before any of this stuff actually works. And uh we kind of endured three or four. Yeah, we endured three or four years of uh of of pretty painful just like sitting in the proverbial garage trying to make this stuff work.

And in 2020, we got the first company to launch avatars as a technology, right? Which essentially is technology that allows you to create talking head style videos like you're kind of watching right now. But instead of recording it to camera with a microphone, etc.

, you simply just type out the script and we generate an AI video of someone saying that. Right.

Back in 2017, everything kind of started with the idea of like creating Hollywood films on laptops and going much more down the kind of creative route of, you know, just like you can publish and write a book from your bedroom. You should be able to do the same thing with Hollywood films. Yep.

Since um since the founding of the company, you know, and kind of where we are today, we kind of took a fork away from that and essentially went into enterprise knowledge. So, what we figured was that let's give it up for enterprise knowledge.

Thank you for they tried to the big Hollywood the creative they tried to pull you. They tried to pull you in. You said no. I am boiled to sex. I love it. I I grew up dreaming about working cross functionally and making videos for enterprise. Let's go. Hit that. Gone. The whole team's fired up. Fired up. Love it.

Making corporate videos for employee train. That's why we do this show. Thank you so much. And it's I mean it's just really smart, right? It's sense. I is clearly good enough to make this form of video and clearly not good enough to oneshot, you know, feature films or even most often even 8second video.

8second video is struggling. Yeah. Um, wait. So, uh, sorry, Jordy, do you have stuff or you you Yeah. I mean, I I I I wanted to jump in briefly because we we spent the first uh half of today's show talking about Sora, kind of reacting to it. I thought it was a great product for entertainment purposes, right?

Cameo functionality. Everything's very memeable, but something about when you're going in the fe going through the feed, it it's it's kind of hard to look at the videos like I have this like very negative reaction. Just something about these these videos. Uncanny Valley.

Yeah, that and then and then the distortions and character inconsistencies is is still rough to look at. I'm curious how you guys specific thing. It's still there's a vibe or feels off.

And so you guys have somewhat of a cap I I'm assuming a lot of the use cases you have somewhat of a captive audience, people that need to watch videos to learn about their role and their companies and things like that, but still you want to avoid people watching watching the videos that you guys generate and having that sort of negative reaction of, oh, this is AI, right?

So how do you think about avoiding that? What what what makes that possible? I think there's fundamentally like two types of video, right? this kind of video you watch because you want to. And this is mostly stuff you'll find in feeds.

So like when you're scrolling Tik Tok, you're on Netflix, you're selecting like what movie to rent, whatever, right? That's something you actively choose to do. Um, and generally you have like a lot of choice, right? Like you could watch like a thousand different action films, but you choose to watch this exact one.

And I think as to your point before, I I think AI will be a huge part in creating lots of this content in the future, but it still feels like the most the reason people mostly watch this stuff today is because it's fun, right? It's novel. It's cool. It's AI. It's kind of weird.

It's not because it's necessarily like amazing content yet. At least 99. 9% of it. There's still some good content out there. I think the space we operate in is much more around knowledge, you know, it's not really around creating storytelling content. It's not advertisements.

It's generally content that you wouldn't comp it's not content that competes for your attention, right? It's content that's very specific and helps teach you something. That could be product marketing as a company, right? You want to teach your pro your prospects why your product is better than the competitions.

There's only that one video to watch, right? And the reason people create these videos is because it's better than text. So the I think on the first category of videos, you're generally competing against like high quality video recorded with cameras and people who really know what they're doing.

H I think the thing we really uncovered back in 2020 when we started building towards the enterprise was that what we're actually replacing is text or slide decks. And in 2025, most people, especially your average consumer, want to watch and listen to that content. They don't want to read.

And for some people, probably other people listening to this podcast, like I'd much rather prefer to like read a page of text and watch an AI generated video. But it turns out that that's not the preference for most people, right?

Um, when people have a problem with their bank and they're trying to figure out like how to do something, they don't want to read a long page of text. If they're trying to figure out how their mortgage works, how much they have to pay in their mortgage, they don't want to read a whole page of text.

They'd much rather watch a video. And so what we're going down, the path we're going down is just much more around like knowledge video, right? It's it's much more of like replacing text with video than it's about replacing video with AI video.

And that's a really really big difference where I think you know when you look at Sara Runway, a lot of these other models, they're much more competing to actually replace like real video with with AI video. And um I think that's also a big market.

It's it's it's super exciting, but we think that there's a much much bigger market actually in in enterprise knowledge and and and turn that into video. And with that, I should also say I think a lot of the talk around AI video is always very much about the models, right?

That's the AI hype at the moment is like there's a new model every second week. That's that's even more impressive than the other one. And that's awesome, right? That's a big component of course of making these technologies work.

But I think we have this mantra internally called utility over novelty, which really is about like, you know, it's it's not about cool demos, it's about bringing value to the customer.

And when you think about why do you make a video, the actual generation of the pixels is a part of that, but it's a smaller part of a much bigger process. Right? So what we've been doing over the last 5 years is really building out a workflow.

We help replace the camera with the AI models, but we also give you a kind of a Canva PowerPoint style editor to finalize your video. We enable someone who's a PowerPoint user, not an Adobe user, to to make their own video collaboration platform. It's a content management system. It's a translation platform.

We have our own video player that's made to work with AI video. And so for us, it's always been about building for the entire workflow of like someone has an idea or some piece of knowledge that they want to communicate to an end consumer. How do we make that entire process as frictionless as possible, right?

And I think what it what it just turns out is that yes, the AI models are really important, but there's so many other components of of making a product that that does this workflow really well, right? I think that that that's what we're much more focused on at Cynthia. Very cool.

and uh very impressed that you were able to to crack 90% of the Fortune 100. That's remarkable. Uh last question, we'll let you go.

Um do you see yourself as being a foundation model company and an application layer company over the long term or is there a world where you uh partner with the foundation model companies to let them you know suck up all the capex to generate frontier models? Do you already do that?

How do you think about the the the shape of the business long term? For sure. I mean, we try to be very intentional about like what kind of company we're building. Yeah.

And I think every company should be and I think it's increasingly clear that competing on the foundation model level is very difficult unless you raise a ton of money. H you hire extremely expensive people. And I think that's just a game that very few companies can play, right? We still train our own models.

I think there's a I think there's lots of use cases where that still makes sense for us. It's again if you work backwards from what are people trying to do with our models, right? It's not just like an 8-second clip for example, it's not very useful.

It's also not very useful that the identity doesn't gets preserved across multiple clips. There's a lot of these like asterisks that you don't see when you watch the cool demo videos of these new models that that are coming out. Um, but we just started also aggregating other people's models in.

So that both means you can just use, you know, V3 inside the platform if you want to create content that's outside of what an avatar model can do. We're also beginning to chain these models together in different ways. We can create a roll and b- roll do lots of cool things with with these models, right?

But, you know, LLM is a huge part of our platform as well. We have a co-pilot product, help you write the scripts, put together the design, the visual.

So, we we train models when we think it makes sense and we think that it gives us a competitive advantage, but you know, we've definitely also aggregated the market and to the extent that we can avoid having to do massive training runs, I think that's the only rational choice to to to do for for a company like us.

That makes perfect sense. Well, thank you so much for taking the time to stop by the TV.