Generating AI videos for enterprise with Synthesia CEO Victor Riparbelli | StrictlyVC London

Hi everyone I am an AI Avatar generated By Synthesia how is everyone doing are You having fun at strict LEC I certainly Am but back to why I am Here I would like to welcome you to a Very interesting fireside chat about Generative AI please welcome to the Stage ingred London writer and managing Editor for Tech Crunch and Synthesia CEO And co-founder Victor riper belly let's Give them a warm welcome and a Round of Applause that was can you guys hear me Okay yeah can everyone hear me yes okay Hey that was not at all awkward was it That little Intro perfect it's the perfect segue for A talk about how natural this AI stuff Really is um so I um just caught Victor Backstage just before this and I said Nothing is happening in the next 24 Hours right that um I need to about You're not going to get acquired or Release some new thing because I mean Really we've been talking about the Answer was no by the way um but I mean It's such a crazy business right now um AI it's moving really fast and um There's so many things happening Including what happened in the last 24 Hours with Scarlet Johansson's voice at Open AI as Mike mentioned um I wanted to Start out with that actually um rather Than talking to you all about how I Expected that yeah I

Well that's good because you work in AI Which is very good at predictive things So yes it's good that you did um yeah I Kind of wanted to just start with that Because it's so funny because I mean you Guys are leaning into it too you had Your last tweet today was or your ex Post um was uh let's see what does it Say it says Sky may not be an option Anymore but hey there's 400 plus AI Voices you can choose from in Synthesia So yeah so tell me what is your uh What's your Takeaway of these things what's your Takea away from that that's your auto Playing video well I I think it's um It's an interesting case right because It is not actually Sky Johanson scy Johansson's voice no um which is an Interesting color I think had they kind Of outright cloned her voice without Permission uh I think it would be for For us at Le very easy it was a pretty Close it was it was pretty close um it Was a clone in the sense that I think They were really going for her voice They did ask her a couple of times for It and she said no I I was can talking Beh of open and what they choose to do For us uh we take concent very seriously We have since day one of of founding Sesia so for us there's a hard line of We do not clone anyone's Voice or visual Likeness without their their explicit

Consent right I think there's um an Interesting non-answer though to that Which I think is that if you look at at AI today it's very clear that the two Best growth strategies is like demo Ware So like make a beautiful video of Something that doesn't actually work When you play with it or it's having no Content moderation and no safety Controls which which help you go viral Especially if you're in like the media Space right and um I think open air is Kind of very good at like both of those To some extent and this is maybe like a Play on the ladder where the kind of Shock factor helps people get talking About whatever isan do you think that They meant to do that on purpose I mean He did tweet her didn't he I don't know I think a lot people decide for Themselves what they think I don't know What what what's what's going on in There I think uh I think everyone who Builds companies especially people who Build AI companies like these probably Like inspired by science fiction I I Definitely am and I'm sure the open ey Team uh like as put your point was was Aware of one of the the interesting most Recent stories around you know what Happens when we can talk to our Computers do you um do you worry though About the way the pendulum might swing Because yeah for sure I mean open AI

Right now is still kind of I still feel Like it's kind Of a bit of a superstar you know um Everyone you know their app is just Absolutely huge um and it's blowing all A bunch of other apps out of the water I Think that you know companies like Google are absolutely scrambling meta to Follow in there to follow on with them Um however there are these little things Happening there was the Scarlet Johansson incident there was that thing With um the the team that was meant to Be running all of their kind of ethical Kind of considerations has been Disbanded you know um do you worry about How that you guys are doing something Very different of course at Synthesia They're Focus fusing on Enterprise um And you guys are building basically um AIS to uh do kind of Enterprise internal Videos um and customer service and that Sort of thing but do you worry about how That is going to have an impact on Companies like yours you know you're Doing something smaller yet very very Connected to that sure I mean I think There there are many Battlegrounds in AI Right I don't think there's just going To be just one company my my sense that Companies like open AI they want to be The infrastructure layer they want other People to build on top of and use their Intelligence to build awesome things

We're very different we're very applied Company we build an AI video platform For the Enterprise where they can make Video content for marketing training Learning sales um whatever and we use Their product right I think right now Everyone is like obsessed about models We talk a lot about technology we have The last 12 to 18 months I think once The kind of hype train sort of slows Down a little bit we'll get back to Talking about products that move metrics For where was buying those products and Solve problems right and and open AI I'm Sure will solve lots of problems for Their customers Will solve a very different set of Problems and I think that that's Ultimately what I'm more focused on like I'm not in control of what a open AI Does or doesn't do but I think from the Looks of it right now yes I mean they're An amazing company right but and they Have a lot of modalities but even you Look at the the image generation right Del is I don't have any stats on this But I'm pretty sure mid journey is like Both in terms of quality but Al in terms Of usage blows Del out of the water so I Don't think it's a mid journey of course Is has less of a consumer Focus than Dy To yeah I don't know I I I me the reports about their their AR Right is pretty substantial of course

Opening has substantial Revenue as well But I'm pretty sure that most people use Image generation as part of an actual Workflow will use mid Journey not Delhi Right and same thing with the text to Speech devices which have also been out For a while now I think we have a great London company 11 Labs who's doing well On that and I think it's they're a Partner of yours they're also a partner Of ours and I think it's It ultimately It comes around so like building great Products right and in our case and I Think 11 labscape that means building a Platform that supports the workflow Around making video or voice content um That has collaboration has Enterprise WR There so many other things to a product Than just the model but because this is So new and so shiny we're all talking About like parameters and models and Technologies right do you feel that Pressure though as a business like so as A startup right now um in AI right now It's very hard to know what the kind of Long-term Horizon is going to be for What a company like yours is um it feels Right now that there's a lot of focus as You say on on this idea of foundational Models these kind of platforms that are Um being developed out of large language Models um are you getting that kind of Pressure from either customers or Investors or internally to build that

Kind of stuff yourselves um do you have People within Synthesia that are sort of Arguing for that or do they all are they Happy to kind of just build on top of What other people are building the way That you are we do very deep AI research Right we have a research team of of Almost 60 people and we're training Foundational models and I think it's for Us it's just you got to have a strategy Right and for us it's like we know the Very narrow slice of a that we want to Own is human presenters in videos in Corporate context which means sales Training marketing right I don't care About Hollywood I don't care about uh Making a video of a dog running on the Beach we'll probably integrate with like Sor whatever comes after that but that Real narrow slice of the world I think We can be uniquely the best in the world At and so that's the only thing I care About and the reason I care about that Is because that's what our customers Want from us so I think looking too much At like you know what other companies Doing in Foundation model is like listen To the customer and what they want and I Can guarantee you that 98% of people who Use these products not just to demo it On Twitter but actually for something Real in their business they don't care About parameters they don't care about Fincial models they care about some

Little random thing that uh that that is Not very exciting but it's very Important to them so like I was actually Just doing uh earlier today looked at we Did we the last 12 months we did 65 um kind of quality of life we should 65 quality of life improvements to the App right which is very boring stuff That nobody here would ever care about But those things you cannot forget even Though the hype train is is going at Full speed right now right it's about The customer it's about what they need And what they want to to solve a problem Do you think though um that um do you Worry about so I look at companies like Open AI or anthropic and they've raised Just ridiculous amounts of money you Know inconceivably billions a lot of it Going towards compute of course um but There's a lot of Hope and you know Valuation writing on those companies They're inevitably going to be moving Right into where you guys are um how are You you don't think so I mean I don't Know that right but if you look at the Big cloud providers like technically the Cloud providers have a lot of like the They have like U trilo functionality Right or like most other kind of SAS That today is like just part of any Company stack they have some version of That product but they don't care deeply Enough about to make it a really good

Product right so I mean there's of Course you see them as like the cloud Company Yeah I think right I think they want to Own the infrastructure are and again I Think even just the act of making a Video right you know make a video in a Chat box that's cool if you're like Making a demo to show your friends but You're actually trying to create a real Piece of content that you need to use For your business yeah so I I don't Think that's like U I think they'll own The intelligence ler for us I see that As as an accelerant um I don't see it as A threat really uh of course I may may Very well be wrong in 10 years time and They've swallowed up the entire world Because they manage to build true ADI They can do anything but but buying that Kind of happening I think uh yeah I Think I think they're they're got to be More like a cloud provider they're going To be every single application yeah okay That's interesting so actually I just Want to go back to something you said a Little earlier you said we don't really Want to you know cater to Hollywood or Anything like that but I think that um So Synthesia has a very interesting Background because you're not like the Technical founder you've got you've got A couple of people who came from Academia who came into this building um

These um kind of video AIS and you've You've come together to build out this Company um which is great um I believe One of your um I think maybe U one of The co-founders I believe did want to Focus on that I think that one of the Original ideas you guys had was about This idea of um you know catering to the Kid in the room who wanted to make a Movie um didn't have the you know Equipment for it I I can't remember the Exact quote Yeah but um and and the idea Is that you were going to build Something that they could use to make Videos it was very much about actually Entertainment creativity creators and so On um do you think you guys will kind of Come back to that eventually or have you Really walked away from that to focus Just on Enterprise I I wouldn't leave that out Completely right but I think we did so We started the company seven years ago So we've been rebuilding this stuff for For quite a long time right and I think What we did you started looking Originally in that sort of thing didn't You or never well I think we I think we Did what you're supposed to not do right Which is we actually fell more in love With the technology than we fell in love With like a very specific customer Problem itself we could solve and that Did the kind of what we anchored

Everything on was like in 10 years You're going to be able to make a Hollywood film on a laptop without Getting anything else in your own Imagination right which may well be the Case that probably will be the case I Yeah but you guys won't be powering it But back then right gen was not very Cool it was very very difficult to raise Money we got turned down by absolutely Everyone in the world and so we had to Seek did you pict it as generative AI at The time we called it synthetic media Unfortunately I remember I had a call With with a bunch of other people Building these Technologies like very Long time ago and we were all Collectively deciding if we should call It generative AI or synthetic media Synthetic media um that would have been Great SEO Jews Today but but what happened right was Just that we sort of went out and we Actually started off trying to build This like doing product for Hollywood And it wasn't a disaster it was did Relatively okay given the company size Right but we just learned that selling To existing video professionals was not Interesting it was a vitaman not a Painkiller and actually there are Millions and billions of people in the World who were desperate to make video But they can't today let's find those

People and actually we give them a Solution like ours which especially when We launched it back in 2020 was way Worse than it is today if it's a Thousand times easier to use and a Thousand times more affordable they'll Be okay with the quality going lower Right and so we identified the first Real use case in the Enterprise for Internal facing content instructional Videos um for where the technology was At the time but our thesis has always Been that ultimately The Tam of AI video And AI content in general is rate Limited by the the quality and the Naturalness and the capabilities right So today the technolog is much much Better we seeing a huge push towards External phasing marketing content we Release our foundation models probably In in three six months time that's going To be like 10 20 x what we have today Maybe you know we put that out into the World uh we're plg type of company right So people come to us and they try all Different sorts of things if we see that People begin to create more storytelling Content like let's see where it takes us I think we're always driven by building A business and doing something that our Customers find useful not just cool Right that that's the key thing for us So if we see people actually using some TCF for real business purposes whatever

That might be that'll be a signal for us To potentially go in that direction but It's very much anchored today on helping The Enterprise communicated video Instead of text right that's the kind of Main idea are you talking much with um Like the other form of B2B business Which is like interfacing with B2B Toc Like are there are there companies that Are approaching you guys so I can Completely imagine the internal video Thing and funnily enough I think part of The reason why it really was a hit not Was just because of the Gap in the Market because there wasn't anything Like that already but because I suspect That the the kind of like uh demands of Quality are so high for professional Video creators or even people who wish They were professional like filmmakers You know the kind of wannabe you know Next Nolan Christopher Nolan or whoever Sitting in his bedroom or her bedroom um They those you know it's just not good Enough yet you know it has to still kind Of do a several more iterations and Today I think what many people Misunderstand about our business is that People look at the videos think like oh They're they're not as good as a real Video why would I watch this video right Yeah um but the key thing is that the Alternative is not text right what we're Doing is we're turning people's text

Into videos so most of the videos McDonald's makes to train all their Workforce globally assum trains their Salesforce with is content which Otherwise would have been a knowledge Based article or piece of text that Nobody reads and nobody remembers now It's a video right so it's more that Kind of Market expansion like I don't Really I'm not so interested in the Video production Market as kind of a McKenzie consultant would would sort of Analyze it much more interested in text I think that's the real market like the World is moving from text to video and Audio and and I think that's that's the Real Market ship that's interesting so Is there an opportunity or are you Seeing one yet for B2B to see there so I'm talking not about like uh McDonald's Kind of you know yeah pushing out an you Know something in their app going hi Have you tried our Big Mac you know I Don't mean that but I mean they would Never do that never um but I mean um I Mean like I don't know talking to a Tik Tok or to snap um you know Snapchat is Really keen on doing more with AI they Have really blown their budgets with it You know they got slammed in the markets For doing too much with the AI that They're so keen on so you know it makes Perfect sense to me that they would work With companies third parties to build

Things out they obviously have done tons With you know filters and so on so There's there's some you can see kind of Opportunities so are you talking with Companies like that yeah we talk with Them all the time and I think Avatar is Is the kind of is this is sort of Starting point right I think for us as Are you doing that already are you Building avatars for anyone or looking At it we have an API people can use Right so if if you want avatars in your Application you can come talk to us and Is anyone doing that yet or we have Customers doing that yeah anyone that You can name any big big ones is it none Of the big ones okay okay may come soon But for me I think like I think the real Interesting thing here is if we're Really successful as a company then in The near-term future Right video will Have evolved from being an mp4 file yeah Which is the file form we use a video Today which is like a sequence of frames That's static and heart baked to Something that's much more Dynamic Interactive personalized and interesting Right okay um and I think that will be a Lot of consumer experience that we'll Move into everyone is talking about like This like now the the hot word of the Week is like agents right agenic Workflows and those sort of things I Think a lot of a lot of that stuff will

Also want be powered by video um again The matureness of these Technologies I Think is the key thing and for me as a CEO what I think about all the time is Like how do you sequence your way where You build a great business right I think There's a a big especially in the Enterprise right now there what I call The Holy Grail problem which is that Every executive you talk to can imagine The end state of AI right like oh I'll Trend LM on all my company data and it Can answer customer requests it can Teach my Sal Team how to do things do All these different things right and Everyone's like cool yeah that that Feels like it's within reach when you Try out like chat gbt but then real life Hits right and and you read all these Walls where like it doesn't really work And the Safety and Security concerns and Maybe the quality isn't where you want It to be what everyone has a very hard Time doing is sequencing their way there Right maybe that's the case in like x Amount of years that you can do that but What does the first step for you look Like what does the Second Step and the Third step and that's what I always Think about like you want you want to You want to Anchor it to something That's valuable today and then people Can use today and get value out of and Then you want to build on top of that H

And I think right now it's just like you Know these videos they're very practical They're very utilitarian it's much much Much better if a customer is trying to Understand how to use your product that You show them with a screen recording Made on our platform and a long wall of Text but no one goes home today after a Long day at work and says like I'm going To open up YouTube and watch a 20-minute Avatar video about something interesting Because the quality isn't there right Later in the year I think we'll open up Storytelling because the avatars will be It'll be so much more better because They're getting smarter or something When you say storytelling what do you Mean so so storytelling I mean um I Think audio has kind of had this moment Already right where like with the with With generated Tex to speeech people are Now okay listening to an audio book That's generated by synthetically right You would not listen to an audiobook From Siri because Siri talked very much Like this right you go crazy after 10 Minutes but it's fine for navigating Your GPS the car because it's better Than like looking at your phone all the Time right and so the quality gap just Needs to get a little bit better and I Think I think we have it more or less Ready internally now actually just kind Of wrapping up and getting ready to ship

It where you know you give it an Interesting piece of text and it can it Can perform that like it's a a very Enthusiastic well performing YouTube Blogger right today it's not a YouTube Blogger it's like a very corporate Spokesperson right right and so that's What I talk about like you know the Tam Kind of increases as the quality the Script it won't be able to kind of Create its own I mean we have those Features as well right of course you can Use an LM to generate your script but I Think it's just I think I again is a ho Problem like oh we wanted to like Generate everything it's like no if you Want to make good content you want to Have a human that looks through it maybe It gets we we think of the llm driven Script and video Generations like it'll Take you 80% there give you a draf but It won't finalize the thing for you Right and maybe there's atic stuff is Then when it becomes like a conversation You can talk with ET I just think there Are many steps along the way before Before that's not just like uh oh my God That's so cool and this is like actually I would prefer to like talk to this Virtual chat agent rather than like Picking up the phone and calling that is A really really high bar and I think I Think a lot of companies trying to build The stuff now will get very disappointed

When they put it on the real world yeah Um I have to ask you one last question Are you talking to Scarlett Johansson we're not talking to Scarlet Johansson though are you speaking with Any I do you see a future for like Adding high-profile people into the into The mix I mean Amazon was doing that Tons with Alexa for a While it's not actually unknown although It's highly controversial you know There's a massive strike in Hollywood Over over AI likenesses so I mean look It's obviously going to happen we are Also talking to we're actually working With some I can't disclose which but Obviously it's going to happen right Like celebrities will want to will want License out their likeness and do more With with their IP some will be very Against it for sure I will never do it There's a lot of people who are very Kind of forward looking in that regard Um and and will will definitely want to Do it so I I think again it's just a Question of when is it good enough it's Not good enough yet today and I think we Just have to be realistic about that Okay I have one last question I'm sorry You guys raised money last June billion Doll valuation congratulations um are You where are you on your Runway because All of the stuff does cost a lot to do We have a very very good business uh you

Know we've always been focused on Building an actual business that's not Held up by by Parachute of VC money so Um I think you know we're good when we Find a way that we think we can deploy Like a lot Capital yeah will will raise But it's not something that's like in The near future okay I'm just going to Ask are there any questions in the Audience hi yes um so thank You hi I'm Monica I have a company which Does realtime mental health support and What you were presenting of this of Course great thing for therapy how far Are we from having digital therapists do You think great question I I think Uh I mean I don't work on the int Intelligence layer right so you're Probably better at telling me when the Actual llm can like provide good and Accurate answers terms of like the video Part of it I don't think we're that far Away really um real time generation is Probably required for this I think the Interesting thing will be as bit as to My point before the bar will be really High for that not to be an annoying Experience right so if you like talk to Your therapist and then it starts Talking and then you try and interrupt It but you can't interrupt it That's really bad right if the latency Is significantly higher than like a zoom

Call your brain will go like I'm so Tired of this just give me the text Version of it um but I think by the end Of the year probably you'll be able to Build something that's like pretty good Uh I don't think I don't I actually I'm Actually more like uh more curious about Like where the intelligence layer is Than the like visualization layer of Having I mean it's quite interesting Because how much are you guys looking at Doing interactive video avatars in General maybe not I mean I think that Would need the highest level for all of The obvious reasons but so we uh we'll Make a big announcement that like in in In the not so distant future as I said Before I think not in the next 24 hours Though not the next 24 hours but but I Think you know for for for all this Right like right now we're at like we're Better way to create MP4 files right and We've proven that's extremely powerful And it's a fantastic business for us but The real like interesting stuff that's Going to happen with AI video is that It's going to become a new media format That's going to be very different from Today and one thing is imagining that And makeing like interactive player cool Yeah that's that's to be one of this Stuff right but where you really want to We need a new type of video Infrastructure for the internet and

That's actually what we're building we Don't see this as just like an Interactive player with a button you can Click on you need to rethink how you Render videos right if you have a Million users interacting with your Interactive video you don't want to Render a million videos on a server Somewhere that's going to get Catastrophically expensive to do so a Lot of the stuff will move to be more Client Side Video will take on an entire New dimension so so we're definitely Working on that but I think it's again It's like the sequence very very Interesting okay okay another question Any other questions Um Can are we allowed to are you allowed To ask a question no let's have you I'm So Sorry thank you hello hi uh my name is Shang and I'm the founder of a company Called ap. Market uh my question is I Have two small questions the first one Is how do you build a moat when AI is Going crazy especially LMS and diffusion Models and everything there's a Google Gemini which happened last month which Is is going to had a Vivo I think which Is a text to video generation pretty Good one for like 60 seconds video open I is going to release Sora how to you Build a mode because everybody in the World can pretty soon generate any kind

Of video that they would want to Especially the ones we saw a demo of and Convert any kind of text or any article Into actual video and the second Question is why are you training your Foundational models I have never seen Anybody do that besides like big Tech Like Amazon Google and everybody else And what does it uh provide to you like Why you're doing that basically those Are two questions I have thank you There's a lot of questions I'll try and See if we can remember all of them yeah Um first of how do we build a mode I Think the same way you've always buildt A mode right like I think if you're if You're planning for your mode to be Training at llm with u a little bit of Proprietary data that's not a mode right That may be a temporarily advantage in Terms of like your goto Market strategy But I think you want to build a you want To build a product that's an integral Part of your your customer infrastru Structure um you I think there's I don't Think there's like any new rules there Right you have to rethink the technology Layer and llm unless you're open AI is Probably not going to be your your your Mode right that even go as far as Arguing like even with the world and how It looks today there's a lot of good Like open source models that like you Know getting there in terms of of that

Right so you want to build login like You've always built login um you Mentioned SAR V all these things yeah They're awesome right but they are not In any shape way or form controllable Um and I think uh yeah they'll they'll Work towards that but I think if you if You ask the people there it's very very They buil for something completely Different right they built to be a Extremely creative ability to create Anything but like consistency of Characters instructing it to create a Video of someone talking I'm not saying That they can do that if they wanted to Of course they can but I think it's very Far down the road map of what they're Working on um and I think that like for The very narrow slice of creating videos Like this I actually do think we can With with the team that we have and you Know we've we've scanned thousands of People in a 3D Studio over the last two Years I think we have a data set that Will uh at least give us enough an of an Advantage right um but I don't think you Should believe about all these Technologies are not going to be your Mode right it's a temporarily advantage In terms of how quickly you can go to Market and that creates winners we've Seen that with open Ai and Gemini most Recently um but the mode you build in The traditional

Maners okay yeah that that might be it Sorry yes don't have any more time for Questions so okay thank you so much Thanks everybody yeah thank you very Much and ing

Coinbase
OUR TAKE

Coinbase is a popular cryptocurrency exchange. It makes it easy to buy, sell, and exchange cryptocurrencies like Bitcoin. Coinbase also has a brokerage service that makes it easy to buy Bitcoin as easily as buying stocks through an online broker. However, Coinbase can be expensive due to the fees it charges and its poor customer service.

Leave a Comment

    • bitcoinBitcoin (BTC) $ 66,540.00 0.68%
    • ethereumEthereum (ETH) $ 3,595.43 1.01%
    • tetherTether (USDT) $ 0.999024 0%
    • bnbBNB (BNB) $ 608.37 0.34%
    • solanaSolana (SOL) $ 148.32 2.4%
    • staked-etherLido Staked Ether (STETH) $ 3,595.52 1.08%
    • usd-coinUSDC (USDC) $ 0.999868 0.02%
    • xrpXRP (XRP) $ 0.488865 0.28%
    • dogecoinDogecoin (DOGE) $ 0.136061 0.12%
    • the-open-networkToncoin (TON) $ 8.03 0.42%