Who Owns a Voice in the Age of AI? with Dan Jasnow

Orlando Wood (01:57.4)
Dan jasnow. Gosh, I'm excited to have you on. It's you know, I've known you for quite some time since the Warner accelerator. You've got a great new job at 11 labs. Thank you so much for coming on the podcast.

Dan Jasnow (01:59.043)
Jazz now. Yes, exactly.

Dan Jasnow (02:13.676)
Thanks for having me, it's great to be here.

Orlando Wood (02:16.126)
It's a real pleasure. I'm really, I always really enjoy talking to you. So I'm glad I get to ask you the good, thing about this podcast is I get to ask like really deep dive questions and you're stuck. You have to answer them. This is great.

Dan Jasnow (02:27.608)
Unless I slam down my laptop.

Orlando Wood (02:30.094)
Yes, exactly. There's storming out is always an option. So listen, I think that like I'm so excited. Like, you know, you've just gone into this new job. IP and legal at 11 Labs. 11 Labs is so rare to me in that it seems to have like its own lane. Obviously, it's, you know, language based model and AI and all those things. But, know,

Dan Jasnow (02:32.216)
Yeah, yeah.

Orlando Wood (02:58.486)
Eleven Labs is arguably the most important voice company on the planet. What is it? What's it feel like to be there? And, know, and, and yeah, what's it feel like to be stepping in there?

Dan Jasnow (03:09.622)
Yeah, I mean, it feels very exciting. So my first day was at our company offsite when, you know, at the time there were about 300 of us. Everybody was together in one place despite being a remote first, very global company. And, you know, that was a really special experience because the energy there was just very palpable. And, you know, there is a sense, I think, among everybody that, you know, we are really sort of on the building state of the art, right on the

Orlando Wood (03:30.37)
Mm-hmm.

Dan Jasnow (03:39.766)
leading edge of this technology. There's a tremendous amount of excitement, both internally and I think that we hear about from our customers that what we're building here is just something that is really unique from a technological standpoint, but also very unique internally. The culture that's been built by our co-founders, Mahdi and Piotr, and that they've been able to sustain through this, continuing to sustain through this tremendous period of growth.

Anytime you're in one of our offices, you we all get together as a company. You can just really feel the energy from everybody and the level of excitement and the pace. I mean, you feel the pace, you know, whether it's from our researchers or our sales team or anybody, just everybody is really pushing and rowing in the same direction. So it's a very exciting place to be.

Orlando Wood (04:34.926)
300 people still feels like startup territory. mean, in my head, the open AI's anthropics have gotten so big and you exist very much in that pantheon of rare companies for me, but 300 people, like that's exciting.

Dan Jasnow (04:46.316)
Yeah.

Yeah, so we were 300 people in late September. We now, we've grown quite a bit since then and we are continuing to expand. Our talent team are truly the unsung heroes of this because the amount of people that they're onboarding is truly remarkable. And we are continuing to add people at a tremendous pace. But we do occupy, as you said, I think a really unique space, which is we are competing against

Orlando Wood (04:53.132)
Right? Yes, exactly. didn't. Right.

Dan Jasnow (05:19.542)
some of these true behemoths and big tech companies that have added generative AI tools to their already existing very large portfolios of services and technologies. But we are still a pretty small player. Our founders always talk about small but mighty. And that also permeates a lot of our culture in terms of we are a very, I think, still flat

Orlando Wood (05:40.323)
Mm-hmm.

Dan Jasnow (05:47.636)
company in terms of structure and you know people do really there's a lot of interaction among our different functions, teams, there's a lot of you know I'd say just strong communication between those teams. There's still very much a startup mentality even though I think we have been able to establish something that is certainly technologically right on par with some of these companies that are much much larger than us.

Orlando Wood (06:17.624)
Yeah, mean, you know, undoubtedly. mean, I think that, you know, it but it also seems to me and maybe this is just an outsider's point of view that although everybody is integrating voice into their tools, you guys don't seem to really have a competitor. You seem to be far ahead of any, you know, voice focused, you know, AI companies.

And you don't seem to have a competitor. Are there competitors that you just keep squashing so readily and easily that we don't know about? Or is your hyper focus in that one area kind of a competitive edge?

Dan Jasnow (06:48.247)
Yeah

Dan Jasnow (06:53.333)
I mean, I think it's certainly true that our hyper focus there has...

allowed us to build an unenduring lead in terms of our research and our product cadence. So that is certainly true. And I think, you you see that across the board in terms of the quality of the technologies, the voice and other audio technologies that we've been able to roll out. Those continue to, I think, set the bar every time we have a new release. But there are, you know, it is definitely a competitive space and, you know, open AI,

Google, know, all the big players are all trying to build these tools. So whether it's...

you know, text to speech, speech to text, you know, other models. There are a number of players out there, but certainly, you know, I think we would say that we have been able to maintain that lead, and part of that is because we've been able to continue to push the pace, and we just have, you know, I think an extraordinary and also pretty small research team that has been able to continue to iterate and to be able to achieve new breakthroughs with every one of these models.

Orlando Wood (08:04.419)
Yeah.

Dan Jasnow (08:04.713)
that we've been able to release in every version of the models.

Orlando Wood (08:07.906)
Yeah. And do you find, I mean, do you find that in some cases, those companies, instead of developing more to keep pace with you and spend burning all that money, is it probably just better to just call you up and go, hey, listen, can you just handle all our voice stuff? And we'll handle the language model training and all that kind of stuff. But you do voice. You do it better than we do anyway. We'd rather just work with you.

Dan Jasnow (08:31.903)
Yeah, so we have a very, I think, balanced business in terms of our enterprise business and our direct-to-consumer business. And so we are doing a tremendous amount of enterprise work.

that's a growing percentage of our business. And that's across the board. we are increasingly working with, if you look at any of the Fortune 500, think our list of customers is very well represented in there. And some of those are public, right? So we...

Late last year announced a major partnership with Meta where we're doing a lot of work for them. We've announced partnerships with other players like Deutsche Telekom.

You know, so certainly, you know, we are seeing a lot of adoption of our technological capabilities, whether that is through public partnerships or non-public partnerships. There is a very clear signal we are receiving, I think, from the market that this technology and our capabilities are definitely...

the market leading capabilities, right? And so even when you have these companies like Ameda who are able to spend tremendous, really truly unlimited amounts of financial resources in order to build these technologies, they're still coming in and saying, we would like to work with you on some of these products. that is an honor for us, obviously. They're a great partner for us. And they have a tremendous, their own tremendous technological offerings. But perhaps our advantage is just that we

Orlando Wood (09:50.99)
Yeah.

Dan Jasnow (10:10.581)
are sort of very singularly focused on voice and audio, whereas Meta is obviously devoting a lot of resources to a lot of different models. But yeah, mean, I think, and it's not just Meta either. I mean, it's across the board where we're seeing that our capabilities, our offerings are best on market, best in market.

Orlando Wood (10:14.882)
Mm-hmm.

Orlando Wood (10:31.342)
Yeah. What are you seeing? mean, I, you know, I want to get into your background because obviously you, you, you occupy a very distinct position. You know, you were, you were at Argent, aren't Fox shift, you know, working in IP, and, and law, you know, you now you're, you know, so you've gone there kind of, you know, sorry, you went from leaving the, an AI industry group at aren't Fox ship.

to being inside one of the fastest growing AI companies, what changed when you crossed that line? I'm sorry, I don't mean to put cross that line as though it's a pejorative. mean, like, you know, when you, when you went from being outside monitoring these things and, and, and seeing them to inside actually directing policy and being on the front lines of change, what changed for you?

Dan Jasnow (11:07.35)
Yeah, no.

Dan Jasnow (11:19.563)
Yeah.

Yeah, so as you said, I practiced in the IP group, primarily trademark and copyright at Aaron Fox Schiff. I was there for 12 years. I led our AI industry group. And in that role, I was doing work with a very broad cross-section of companies, primarily advising them on what it means from an IP standpoint to be integrating AI technologies.

with their portfolio of content in whatever systems they had, to be using it to develop new content. But I was also, that includes great companies like WBD that I worked a lot with, but it also included companies in a lot of different industries.

And it included both sort of AI native companies and non-AI native companies. So in that role, think primarily at Aaron Fox, what I was doing was working with a lot of different areas, seeing from a 10,000 foot level, how are companies in different industries adopting AI?

What are the use cases that they're focused on? What are the main guardrails that they want to have in place? What was important to them from a design, architectural, implementation standpoint? And what kind of policies did they need to put in place in order to manage their employees or their contractors or anything like that? The great thing about moving in-house to a company like Eleven Labs that builds its own proprietary models is that it gives you a chance to really

Dan Jasnow (12:59.713)
go deep into how these models work, what goes into them, what are the guardrails that you can put in place on the developer side, what are the IP protections you can put in place on the developer side, and then how do you build partnerships with the type of companies that I used to work with in a way that both parties feel like there's good value being generated. And so I think...

Orlando Wood (13:18.296)
Mm-hmm.

Orlando Wood (13:24.867)
Yeah.

Dan Jasnow (13:26.773)
What has been most exciting to me in this role is being able to get the kind of front row seat on the developer side, the personal education that I've been able to get here about just how these models are built, how they're operated, who's building them, the type of guardrails we're putting in place.

know, deepen my own understanding about, you know, sort of that technology. And that's something, even though, you know, I was doing some work with 11 Labs in my former job, it was the front row seat that you get in-house is just completely different. And it's been, you know, very, you know, personally, you know, invigorating for me just to be able to, you know, have that opportunity to learn these things and see it from this perspective.

Orlando Wood (14:00.366)
Mm. Yeah.

Orlando Wood (14:13.804)
And so what changed most in your outlook? know, governing these things and talking about IP and legal frameworks now being in there, what's a misunderstanding you had before that you don't have now? Or how has your relationship to your understanding of it changed?

Dan Jasnow (14:37.108)
Yeah, I don't know that my understanding has changed that much, but I think I've gotten, it's refined my perspective to some degree. And I think the biggest area in that is...

that every model and every developer is unique, right? And everyone sort of has its own flavor. It has its own risks and guardrails. But there's a lot that you can do with these models and whether that's sort of front-end engineering or whether it's after the fact protections that are put in place, moderation.

red teaming, all sorts of things that you can do in order to make sure that you are developing and deploying models in a safe and responsible way. And I think, particularly in the media and entertainment space, there's a pretty wide misconception about how these models operate and about what the level of control can be. I think we've seen that a lot recently with the Seed Dance, Tom Cruise and Brad Pitt video that

Orlando Wood (15:27.916)
Mm-hmm.

Dan Jasnow (15:48.163)
got a lot of buzz. That kind of thing, I knew about this before, but I think particularly from my perspective now, being an in-house IP council at a developer is just that...

You know, there's a lot you can do on our side to make sure that people are not using content without authorization. They're not using name, image, likeness, or voice without authorization. So, you know, when there is something like that that comes out that, you know, think, you know, validly causes a lot of consternation among, you know, people in creative communities, actors, directors, etc.

you know, and they maybe attribute that to all of generative AI, like this is a huge, you know, fundamental problem with generative AI.

You know, now I think more, no, you gotta look at that particular model. You gotta look at who, you know, particularly, who built that model, who deployed it, what guardrails are in place for that. All of these things are ultimately very solvable problems. And I think something that 11 Labs has been very good at and is one of the reasons that I really wanted to join 11 Labs was because, you know, all of our models and all of our deployments, you know, we've put in a lot of protections to make sure that rights holders have consent, you

Orlando Wood (16:39.939)
Yeah.

Dan Jasnow (17:00.842)
We're working directly with rights holders for a lot of our data needs. We're working directly with rights holders on content distribution, working with people like Matthew McConaughey, directly saying, you know, we will give you access. And Michael Caine. And, you know, that is just that's just a very different model than, you know, some of what I think people have a lot of concerns about. And I think those types of instances create a misconception about

Orlando Wood (17:12.61)
Yeah. And Michael Caine as well. Yeah.

Dan Jasnow (17:30.186)
maybe increase the level of fear or anxiety around these technologies because people think there's not going to be control. But in fact, there are lot of ways that you can put in place good controls and make sure that these tools are being used responsibly.

Orlando Wood (17:44.482)
Yeah, I mean, it seems that you guys are really lent into credit control and compensation. And Matthew McConaughey, especially, has been very vocal about being like, you know, yeah, everybody stole some stuff. That's not great. But don't stick your fingers in your ears and miss this thing that's happening. Make sure you're engaging. Make sure you're owning yourself, your likeness, your voice, all of that. And he's been very vocal about that. And I think that's great. And you guys,

not only have controls in place for that, but also, you know, having used your the product a lot, a lot of controls to make sure that just everyday people can't have their voices taken, you know, and that there are you have to read unique things on the screen to make sure that the person that you're trying to copy is in the room and that they're giving their personal consent. Things like that. Things that a lot of the, you know, the open source models don't do and a lot of other companies wouldn't have done.

Dan Jasnow (18:34.472)
Exactly. Yep.

Dan Jasnow (18:42.378)
Right, yeah, and we have a number of things in place. One you just mentioned is the voice verification, which we have in place for professional voice clones. We have what we call a no-go voices list, where a large number of celebrities and politicians are added to that

and we block generations that are closely similar to that person's voice. And then obviously we have additional moderation and text classifiers that will block certain things that either are high risk from an IP standpoint or violate our prohibited use policy. So yeah, we have really, I think, leaned in and made it an imperative for the company to make sure that we have these types of guardrails in place. And I would say,

You know, there are certainly, that can cause friction with consumers, with users of the tools. anytime you do something like that, it increases the moderation rate. So, you the rejection rate when somebody is maybe coming in and trying to create generative content. But the flip side of that is that, you know, I think those types of protections are essential as we are building out our enterprise business, which, you know, is not just my opinion. You know, there's widespread belief in that within the company from the top down.

Orlando Wood (19:42.294)
Mm. Mm-hmm.

Orlando Wood (19:54.509)
Mm-hmm.

Dan Jasnow (20:00.599)
that as we are doing this, as we continue to partner with major players, major rights holders, major IP owners, it's very important for them and for us to be able to show that these tools that we are bringing to market are being built and deployed in a responsible way. So as much friction as there sometimes is with individual users who are seeing maybe a higher rejection rate with their prompts than they might with some of the other tools that are out on the market.

It also is really essential, I think, for our long-term, for the trust we have built and are very intent on maintaining with some of these large enterprise customers that we have.

Orlando Wood (20:36.503)
Mm-hmm.

Orlando Wood (20:42.53)
Is there a world in which you create a marketplace for all people to upload their voices? I mean, I'm thinking in particular, like it's great to have Matthew McConaughey and Michael Caine. Like, that's terrific. I'm thinking about so many of the jobbing voiceover actors and artists who, you know, who kind of go in and job for that kind of work. Is there a world in which you empower them as well?

not just the sort of the creme de la creme.

Dan Jasnow (21:10.334)
Yeah.

Yeah, so we have that today. There are two main offerings that we have for our voices.

Orlando Wood (21:15.719)
great.

Dan Jasnow (21:21.491)
they're somewhat different. So we have what we call an iconic voices marketplace, which is where Michael Cain's voice resides and Liza Minnelli's voice and a couple of these other and Maya Angelou. Those are some of the most high profile individuals that we have who we have contracted directly with either their representatives or their estates in order to have them be available, have their voices be available to users on our platform. For the iconic voices marketplace, you cannot

just go there and use someone's voice. You basically come onto our marketplace, you submit a request to use the voice, and then those rights holders get that request. They have complete control whether to approve or deny a particular use. So that's...

Orlando Wood (22:04.344)
great. OK. So they're not just handing over those rights to you to manage it. You know, they they are constantly included in that discussion. Right.

Dan Jasnow (22:14.483)
Yes, they have complete control over whether or not their voice will be used. So if the rights holder gives us a thumbs up for a particular referral or request, then they'll come back to us and say, this is OK. And then we might help them produce that. So we would maybe create the voice clone or the synthetic audio, whatever it would be, that the requester could then use. We also have our general.

Orlando Wood (22:32.078)
Mm-hmm.

Orlando Wood (22:39.47)
All right. So they're not selling something. They're not going to get back. I think that's one of the big fears about most artists. They're like, I'm I'm licensing out or selling it out. It's they're not giving up control. They're just they're just giving people the ability to browse and understand what that voice could do for them and then make a decision about, know, you know, if you're Reese's and you want to continue working with Will. dear. I can't believe I'm spacing out his name.

Is it not Will Forte? dear, what's his name? But anyway, it doesn't matter. But if you're creator and you see a voice, you can browse it, but they have full control.

Dan Jasnow (23:20.437)
They have full control. We are never going to put something out sort of in the wild that they haven't authorized. So in the case of our Iconic Voices Marketplace, we see it really as just an opportunity for those rights holders to deploy their rights to...

you know, for additional projects and to derive additional revenue. So if you're Michael Caine, you know, you can have the right to say, I want to allow my voice to be used to narrate this particular TV commercial. And we will produce the audio if he gives us the green light for that. And then, you know, we'll deliver that audio file to whoever he has agreed to work with. And then, you know, they'll use it and then it'll be over, you know, depending on whatever terms the rights holder in that case, Michael Caine agrees to with.

Orlando Wood (24:08.302)
Mm.

Dan Jasnow (24:10.071)
whoever the licensee is. So total control, no replacement in that case.

Orlando Wood (24:13.667)
Yeah.

And presumably that puts you right in the sweet spot for the like SAG after negotiations are going on now. This was a huge part of what they were talking about. Presumably that framework makes them entirely happy.

Dan Jasnow (24:30.185)
Yeah, yeah, absolutely. And you know, we are...

very much aligned with the importance of the consent model, right? I was on a legal panel a couple weeks ago now, where, actually the panel before me, there was a lot of consternation about a lot of these issues that the unauthorized use, and do people understand if they are granting their rights that their name image voice likeness might be used for AI? We don't operate really in that gray area. If you come to us,

We are a voice AI company, and we are entering into direct agreements with these folks. Particularly in the case of the Iconic Voices Marketplace, Michael Cain is coming to us specifically to request our services to create AI generations of his voice. In those cases, it's exactly the model that SAG is talking about, which is very clear, express prior consent, compensation, transparency, control.

Orlando Wood (25:07.79)
Mmm.

Dan Jasnow (25:35.159)
that I think we have been able to implement very effectively. And to go back to your earlier question, this is something that we do make available to anybody. So anybody can come on our platform and contribute a voice to our standard, our regular voice library, and make that available for third parties to come and use. And any time those voices are used by other paid users, then you get compensated for that use.

Orlando Wood (25:38.104)
Hmm. Yeah.

Orlando Wood (26:00.412)
Mm-hmm.

Dan Jasnow (26:04.922)
So, you in that case, if I were to go and contribute, actually I have, you know, for, contribute a voice clone to our voice library, somebody can come on there, I can choose whether or not I want to share it in the voice library, and if I do, you know, there are payout rates that you can agree to. you get paid, you know, X cents or, you know, of cents per thousand characters, and then, you for some people, that, you know, for some people, and it is a minority, but...

Orlando Wood (26:22.531)
Mm-hmm.

Orlando Wood (26:29.443)
Mm-hmm.

Orlando Wood (26:34.67)
Mm hmm. Yeah.

Dan Jasnow (26:34.742)
but we are paying out a very large amount of money on a monthly basis based on use. And that is really just, it's an open marketplace. So if there are voices that people really like or if they go viral on one of the social media sites, they are getting paid for that. And that includes a lot of voice actors and just other folks who have decided to contribute a voice and are now profiting from that in the form of passive income.

Orlando Wood (26:48.344)
Hmm.

Orlando Wood (26:54.595)
Yeah.

Orlando Wood (26:59.726)
And can you do performative voices? mean, could you, you know, if I were a Hanna Barbera voice actor and I had 10 different voices, I do, could I kind of create a suite of my best voices and have them all, you know, up there? Great.

Dan Jasnow (27:14.022)
Absolutely. Yeah, yeah, yeah. have lots of people who have multiple voices on the platform. You know, everything has to be subject to our terms of use and prohibited use policy. So you're not allowed to contribute anything that you don't have the rights to contribute. You do have to verify that.

Orlando Wood (27:27.884)
Yeah, you can't mimic Bugs Bunny, but if you have a unique character that you wanna put up there, you can make that available for others. Yeah.

Dan Jasnow (27:32.221)
Right. Right.

Exactly. And people are doing that. so not only do we have sort of the voice cloning, so you can, in our parlance, you can create either an individual voice clone, which doesn't take as long to record, or a professional voice clone, which takes more like 30 minutes of audio. So that's a clone, but then you can also use our voice design tools, voice mixing, and that's where you could, if you can do a sort of cartoonish character,

you could create a voice clone out of that, and then you can use voice design to sort of tweak it. So there's lots you can do. There's a lot of artistry involved, and there a of tools on the platform that allow folks, voice artists, creators to exercise that sort of artistic sensibility, applying that to this unique technology.

Orlando Wood (28:26.924)
Yeah, you would mention before that there's too much fear and not enough understanding around AI. Where does that misunderstanding do the most damage? Is it in the courtroom? Is it just in the press or just in, you know, in people's understanding at large?

Dan Jasnow (28:42.694)
I think it's across the board.

I understand why people are apprehensive about generative AI writ large. I it is very powerful technology. It is gonna have significant social impacts. I think there's no doubt about that, but it is also something that offers tremendous potential. And it is something that is just here.

just like the steam engine was here, just like motor vehicles are here, just like air travel is here, just like the internet was here in the early, in the 90s when it really took off. mean, this is a fact of life that is here that has both tremendous upsides and a lot of risks. And it's very incumbent, I think, on everybody to understand both aspects of that because it's gonna affect all of our professional lives. And I think for kids today,

Orlando Wood (29:17.87)
Mm-hmm. Mm-hmm.

Dan Jasnow (29:44.839)
You know, kids in, I've got a seven-year-old daughter, but anybody who's coming up through grade school that's of a certain age, this is gonna be essential to their workforce, to their workplace. And particularly if you're a little older, if you're in college today, if you're thinking about going to law school, if you're coming out of law school today and you don't have an understanding of how this technology works, both on the advantages and the risks of it, that is, you

Orlando Wood (30:09.624)
Hmm.

Dan Jasnow (30:14.742)
if I were hiring somebody, you know, a first-year associate, a summer associate at a law firm right now, I would say that's disqualified because, you know, there is a tremendous amount of efficiency you can get.

Orlando Wood (30:22.04)
Yeah.

Dan Jasnow (30:29.236)
And, you know, but I think going back to the misconceptions, the biggest one to me is that this is a way to leverage your own personal knowledge and expertise. So if I'm a lawyer, you know, I can use AI to accomplish, you know, 10 things today in the time it would have, you know, much faster than it would have taken me previously. And that's essential to being able to keep pace at a company like Elevenlabs, right? But that's true.

at law firms or any other company. And that doesn't mean that I'm letting the AI exercise the judgment for me. I'm still going to be looking at everything that I'm producing. I'm still going to be revising everything that I'm producing. But that time savings is an essential...

skill that I think we all are going to be expected to have in the not too distant future. And so I think, you know, you can apply that to anything. Everybody should be, I think, educating themselves on how to effectively use these tools, how to leverage their own expertise, not replace their expertise, not replace their judgment. I still think, you know, people need to be learning critical skills, how to read, how to write, how to be analytical, how to be thoughtful, skeptical.

Orlando Wood (31:43.0)
Yeah.

Dan Jasnow (31:47.878)
But with that strong foundation, there's a tremendous amount that you can do with this technology and that I think is gonna be happening and the people that understand how to do that are gonna be at a much greater advantage.

Orlando Wood (32:00.462)
Yeah, it's funny. A.I.'s getting blamed for a lot of stuff that actually was already happening before A.I. came around. Like I was I was watching some third grade or fourth grade teachers say, you know, no kids know how to read a clock. No kids know what their address is. They don't know. You know, like, I mean, the the amount of critical skills that have died on the vine just from because of our generalized digital life and social media, that's that's been killing a lot of stuff. And now A.I.'s being sort of blamed for it.

Dan Jasnow (32:14.695)
Mm-hmm. Mm-hmm.

Dan Jasnow (32:25.029)
Mm-hmm.

Dan Jasnow (32:30.418)
Yeah.

Orlando Wood (32:30.488)
But I wonder, I mean, is there, we seem to be living in a moment where regulation either hasn't caught up or is kind of given up. I mean, what is going on at the moment? Is there a decision that you're worried is going to come down the pike? Is there a shift in, you know, if the Democrats win in the midterms, is there suddenly going to be regulation? Is there even a hope that

an organization as large and as cumbersome as Congress can even make decisions at the speed of a nimble startup that's 300 people. I mean, 300 people plus now. I mean, what hope do they have? And then, and then a technology that's moving at such speed by the time you legislate whatever, you know, 11 labs, 3.89 that you may have out today. I'm sure I got that wrong.

Dan Jasnow (33:03.174)
Mm-hmm.

Dan Jasnow (33:09.073)
Yeah. Yeah.

Dan Jasnow (33:16.435)
Mm-hmm.

Orlando Wood (33:25.558)
You know, by the time they legislate it, you'll be on 11 Labs 10.8. You know, I mean, what hope did they have?

Dan Jasnow (33:28.211)
Yeah, yeah.

There is no hope. Abandon all hope. Yeah, I mean, and so I would say that as, I think we need good regulation. And I think there's a wide consensus within the industry that regulation is essential. I often compare it to crypto, where I think that market was allowed to, people would disagree with this. But if you look at the failure of,

Orlando Wood (33:34.306)
Right.

Dan Jasnow (34:00.793)
FTX and some of the excesses there, any potential sort of, at least a lot of the hope for widespread adoption of blockchain, of crypto, I think was set back very significantly by widespread fraud or distrust within the industry and the failure of FTX and all that kind of stuff.

And I think, you know, it's critical for something for an industry like AI to maintain consumer trust, to maintain enterprise trust, to show that these tools can be used safely and responsibly, to be good partners with government and to find regulatory partners that understand the technology. I think there is very little hope that that is going to happen at the federal level. I think it could happen, at least in the U.S., at the state level,

some states. That is a very delicate dance. I think it is very hard to strike the right balance between being overly prescriptive and making sure that the appropriate safeguards are in place. But I certainly think that there's an important role for regulation to happen in this space. And I often, yeah.

Orlando Wood (35:23.576)
Well, also Anthropic now just saw that self-regulation is a sales tool. mean, they went to number one after, you know, they, they refused to, kowtow to some of the requests of, you know, the department of war. They hit number one on, on Apple, on the Apple app store. you know, so are you, are you seeing that as an, as an edge actually, I think that like being responsible, I think at least the.

Dan Jasnow (35:28.637)
Mm-hmm.

That's true.

Dan Jasnow (35:38.215)
Yeah, yeah.

Dan Jasnow (35:42.855)
Yeah. Yeah.

Orlando Wood (35:53.42)
the people consuming these products appreciate it deeply because I think everybody's scared on some level. so watching a company say, no, we won't do that. Even if you are paying us 200 billion or whatever it was.

Dan Jasnow (35:57.892)
Absolutely. Yeah.

Dan Jasnow (36:05.329)
Yeah, yeah, I mean, that is absolutely right. And I think that gets to sort of the end. Yeah, 200 million, I think. Yeah.

Orlando Wood (36:09.614)
or 200 million, think it wasn't the big, yeah.

Dan Jasnow (36:14.579)
That gets to enterprise trust too, right? I think there's a tremendous value for us in making sure that we are being responsible in the way that we are building these tools and deploying them when it comes to major partnerships with, for example, a Meta or a Deutsche Telekom or any of our other very large enterprise customers.

And that's, think this year in particular, we're going to be continuing to announce major partnerships, major customer relationships with major enterprises, and particularly in the media and entertainment and publishing spaces where there's a lot of concern within those industries in particular about abuse of IP rights. And so it's taken several years, and I think a lot

Orlando Wood (36:42.488)
Mm.

Dan Jasnow (37:06.665)
of work behind the scenes to establish trust and to show that you know we can build these tools while also being a good partner to rights holders and delivering value back to to individual rights holders whether that's authors or actors or

directors, So, you know, and I think a lot of this flows from steps that Eleven Labs has taken over the years, right? So last August we were first to market with a fully licensed music model. So there are a couple of competitors in the space, Suno and Udio, who both were sued by the major labels, and we didn't do it that way. We went out, we struck some deals first, we licensed content in order to train on.

And then we were able to come to market with a tool that was just as, if not superior in terms of performance, but also was fully licensed. And we are sharing the revenue back with the individual rights holders, the musicians, the songwriters, et cetera, who make that model possible. And I think it's those types of decisions that those are decisions that we are making that don't always

Orlando Wood (37:58.392)
Mm-mm.

Dan Jasnow (38:19.29)
How can I say this? They're not always the most expedient decisions, right? It takes time to go out and to license and money to go out and license this content to build those relationships, to negotiate those agreements. But in the long run, there's tremendous long-term value that can accrue from doing it that way. And I think we will continue to see that. And I think that momentum is going to only accelerate this year.

Orlando Wood (38:23.064)
Yeah.

Orlando Wood (38:27.096)
Mm-hmm.

Orlando Wood (38:41.39)
We saw that we read at the start of, you know, so much of this. We fine tuned a model based on a foundational model that we knew to have no copyright training data in it. We then licensed a bunch of screenplays and then made a fully a screenplay focused language model that that allowed us to to get into a lot of places.

Dan Jasnow (38:53.904)
Mm-hmm. Mm-hmm.

Dan Jasnow (39:01.841)
Hmm.

Orlando Wood (39:04.962)
The unfortunate reality of it was is everybody was like, well, I've tasted champagne. I don't want to go back to beer. You know, like there was, but we got a lot of, we got a lot of trust in the market that we spent the significant amount of money to do that. And that we had that model and you know, listen, lots of clients used it for a few weeks and then went, you know what, move us over to GPT or move us over to Claude, because actually it's a much more functional model. It's built on.

Dan Jasnow (39:09.85)
Right. Yeah.

Dan Jasnow (39:26.586)
Yeah, yeah, yeah.

Orlando Wood (39:32.266)
huge amounts of stolen material and so it does work a lot better. But I think that we did find that as well is that a lot of our clients stuck around because they were like, we like that you cared enough to do it, you know?

Dan Jasnow (39:33.958)
Yeah, yeah, yeah.

Dan Jasnow (39:44.615)
Yeah, yeah. And you know, I think there is...

Certainly something unique about the audio space where we've been able to achieve very, very powerful models. Our models are quite a bit smaller in terms of compute size than if you like the GPTs, mean, orders of magnitude, right? So we're able to accomplish a lot more with smaller amounts of data. And because we have an exceptional, truly exceptional world-class research and engineering team, they are able to produce things that are quite extraordinary

from relatively small amounts of data. And that is something you can't do so much in the LLM space where, you know, for GPT or Claude, you just need unbelievable amounts of data. And, you know, they are operating on a very defensible, fair use basis, as, you know, everybody in the industry is to some extent.

Orlando Wood (40:36.002)
Mm-hmm.

Orlando Wood (40:44.844)
Mm-hmm.

Dan Jasnow (40:46.006)
We will see where that develops. You mentioned what cases are coming down the pike. Certainly, that's something that everybody is waiting for, probably ultimately a Supreme Court decision on when is the use of this content fair use. it'll impact everybody in this space for sure, everybody in generative AI, because there is a tremendous amount of data need.

Orlando Wood (40:58.19)
Mm-hmm.

Orlando Wood (41:01.836)
And does that impact you as well?

Dan Jasnow (41:13.67)
But these companies, all of us are operating globally and so having it addressed in the US will be very influential, but it's also not gonna be the only jurisdiction that's important here. There's a lot going on globally pertaining to this question about how...

how rights should be allocated in the space, the types of permissions or consents that should be put in place, what the right public policy approach is. We're seeing in places like Brazil and India a lot of talk about compulsory licensing regimes for scraping of content. And then obviously in the US you have these questions about fair use. The EU is debating when are there, what are the appropriate exceptions to copyright?

Text and data mining exceptions is sort of the nomenclature over there that's most relevant. So, you know, there's a lot that is under development. Those decisions will be very important. I think for us, you know, more than many of our peers, we have really pursued a licensing strategy. We're able partially to do that because we don't need quite as much data, although we still have tremendous data requirements. But we don't need

Orlando Wood (42:29.71)
Hmm.

Dan Jasnow (42:37.39)
as much as maybe an open AI or an anthropic for sure actually. So I'd say we're better insulated from some of those decisions, but in the long term, how those policy questions are resolved globally is going to have a tremendous impact on everybody in this space.

Orlando Wood (42:40.782)
Mm-hmm. Yep.

Orlando Wood (42:54.784)
Mm-hmm.

Yeah, so I mean, no, it's interesting to hear that you guys are kind of less data hungry than your competitors. And yeah, that's terrific. I mean, is the legal uncertainty actually great for innovation? mean, does it sort of mean that you guys can kind of keep running at speed? And then there comes a point where even if the

decision goes the wrong way and they say, or not the wrong way, but the decision goes the way of deciding that it's not legally fair use. You're sort of, at least you have the ability to have grown large enough that that becomes a burden that's okay to carry.

Dan Jasnow (43:41.202)
Yeah, mean, you know, certainly, you know, we feel pretty confident about how these things are ultimately going to shake out. So, you know, question of fair use. I think, you know, the longer this stuff. Well, I think if you look at the case law, if you look at the precedent over the last 30, 40 years.

there's a pretty clear signal there in terms of how this will likely ultimately shake out. There's no guarantee on that, and that's one of the reasons why we're hedging on some of these other, that it'll be considered fair use. So not maybe in every circumstance. If you look at the BARTs, the anthropic case, that's sort of the one that's farthest along.

Orlando Wood (44:13.688)
And what's that? What's that outcome?

Dan Jasnow (44:26.746)
In that case, the court said basically, if you, for any of these books where Anthropic had gone out and purchased physical copies and then digitized those and used them for training purposes, that was considered fair use. Where they had used torrented books, then that, exactly, right.

Orlando Wood (44:43.778)
or like the bookstreet corpus and things like that, that's maybe right.

Dan Jasnow (44:47.826)
That was not fair use. even there, there's a pretty strong signal for fair use. I think a lot of people in the industry would say that these technologies are profoundly transformative, which is one of the core parts of the fair use analysis. And even the judges, the federal judges in the US who have looked at this have said, I think you've used exactly that language, that it's profoundly transformative.

Orlando Wood (45:16.12)
Woof! Woof!

Dan Jasnow (45:17.96)
But, you know, certainly I think there's also a recognition that, you know, there needs to be, there's a great opportunity for value to be shared, you know, across everybody who is involved and affected by this industry, you know, legacy rights holders.

Orlando Wood (45:34.232)
Yeah.

Orlando Wood (45:37.646)
Mm-hmm.

Dan Jasnow (45:39.204)
And so I think that's reflected in things that we've done like our music, fully licensed music model. So ultimately, you know.

Orlando Wood (45:42.414)
Mmm.

Yeah.

Dan Jasnow (45:49.254)
The legal uncertainty is sort of a fact of life in this industry. We feel pretty confident about where things are headed. you also can't, litigation takes forever. And these technologies are here today. And so for in-house legal teams like ours or like the teams at any of our competitors, you have no choice but to sort of be looking at the latest developments, seeing how various

Orlando Wood (45:52.974)
Mm-hmm.

Orlando Wood (45:58.222)
Hmm.

Orlando Wood (46:01.795)
Yeah.

Dan Jasnow (46:19.277)
legislative bodies are moving forward and to be just adjusting in real time based on the latest signals you're getting.

Orlando Wood (46:24.366)
Well, I think that the other thing is that, know, you guys are, you know, voice is likely going to become the primary way people interact with AI, you know, not, you know, even the things that have been talked about with Johnny Ive and Sam Altman's, you know, tools that they're planning to build or don't have are rumored to not have screams, not have keyboards, you know, that that.

Dan Jasnow (46:49.158)
Mm-hmm.

Orlando Wood (46:51.042)
means that like, you're a you become a voice laver there's layer that sort of everything runs on. That's an incredibly powerful position to be in.

Dan Jasnow (46:56.561)
Yeah.

Yep. Yeah, I mean, that is the North Star of the company. That voice is the way that we will interact not just with AI, but with all technology, and that it is the most natural way for us to be interacting with technology. And we strongly believe, and I think the market bears out, that we have the most sophisticated and...

highest performing voice capabilities and audio capabilities. So we see that as an enormous potential, an enormous opportunity for 11 Labs specifically. But obviously, it's incumbent on us to make sure that we are continuing to press our advantage, continuing to develop, iterate, and press what we believe is our research advantage in this space.

Orlando Wood (47:51.596)
Yeah. And how much, you know, so there's going to be the internet connectivity of all of the devices that we use. And that voice layer is will be very important to that. But also we're heading towards robotics and ambient computing and sort of a voice driven everything. You know, what what are the what are the things that you guys are are focused on on those horizons?

Dan Jasnow (48:03.419)
Mm-hmm.

Dan Jasnow (48:15.697)
Yeah, mean, certainly we are looking at a number of different integrations. I think if you look at any of the hardware that's out in the world that is best equipped to be controlled by voice, think about cars, think about glasses, think about smartphones, obviously. We are...

I agnostic in terms of those different modalities or channels, but we are definitely pursuing, I think, many if not all of them. If you look at also what's already on the market in terms of things, other voice assistants, I think there's definitely still a pretty wide gap between consumer experience and where the technology, certainly where

technology is today. So you know we see a lot of opportunity to be strengthening some of those, improving the performance and making those interactions more natural and that's you know that's across the board. Anything that currently could be integrated with a voice assistant today you know is something where you know I think would really benefit from 11 Labs voice capabilities and agents, conversational agents which we are deploying.

Orlando Wood (49:42.138)
wow. Are you already deploying those conversational agents or is that a scoop? Is that something on the horizon?

Dan Jasnow (49:45.936)
Yeah. No, have, so we recently rebranded, we've been doing, we have been,

We've had an active agents platform for a number of months. So we now have 11 API, 11 agents, and 11 creative as our core platforms and products. so 11 agents is something where our conversational agents are being deployed today at very large scale and things like call centers.

customer service applications, all sorts of different things. those are, the capabilities there are really extraordinary and are things where we're seeing very significant improvement metrics from our customers.

Orlando Wood (50:37.76)
Mm hmm. Well, I mean, yeah, I think it was one of your last releases kind of cracked listening and talking almost in real time. You know, obviously one of the big bugs of voice agents and and voice interaction was there was this meaningful pause after you finished speaking and then it would start replying to you. Now it's replying almost in real time. I imagine that for customer service and robotics.

Dan Jasnow (50:56.057)
Mm-hmm. Yeah.

Orlando Wood (51:06.37)
That's a huge leap forward and it doesn't seem like anybody is close to that ability yet.

Dan Jasnow (51:06.703)
Right. Yeah.

Dan Jasnow (51:13.039)
Yeah, yeah, absolutely. The learning of the turn-taking, the natural turn-taking in human dialogue is something that the models are getting much, much better at. And it's something that is reflected in our models. There is, you know, and there will continue to be, you know, latency issues. You know, it's just, and sometimes there is a, you know,

there's a little bit of a give and take between latency and sort of performance. you know, if you have the lowest latency is going to have, you know, maybe.

little bit higher performance issues, maybe not, but either way, know, there's always going to be a little bit of a give and take there. But, you know, it's something that we have been able to, you know, continue to be sort of a leader in, you know, lowest latency models, really exceptional performance across a very large number of languages.

Orlando Wood (52:02.318)
Mm.

Orlando Wood (52:09.568)
you cross that uncanny valley, you know, surely you guys have done, but that response...

Is there a moment where it becomes false again? Like, is there a moment where you go, well, actually in trying to make the customer service AI rep sound so human, actually like now, like the bug was almost a feature at some point, because at least people knew they were interacting with an AI. Now there's almost no ability to differentiate. Does that matter? Is there a

Dan Jasnow (52:40.58)
Mm-hmm.

Orlando Wood (52:46.816)
A legal or ethical responsibility on your guys end to let people at least know they're interacting with an AI model.

Dan Jasnow (52:54.574)
Yeah, and I think we already see that in some of the regulatory landscape. So there are a number of states who have adopted laws who say that in certain, least, regulated domains, you need to disclose if a consumer is interacting with an AI agent, and particularly if the AI agent is making a material decision, an automated decision that could have real...

financial or other material impacts on that consumer. You need to disclose, you need to provide an opportunity for a human to be involved. So I think part of that is a policy question. It's one that the state legislatures have already started to think about and to, in some cases, legislate on. And I think that's important. I always go back to sort of a fundamental principle of consumer protection law, which is, is the interaction misleading or deceptive?

Orlando Wood (53:37.934)
Mm. Mm.

Orlando Wood (53:49.368)
Mm-hmm.

Dan Jasnow (53:50.237)
misleading the consumer, deceiving the consumer, or otherwise engaging in some sort of unfair or harmful practice if there's not a disclosure or if the agent is engaged in some sort of automated decision making. So I think there's an important question there. I think that's an important policy question that should be addressed and that would benefit from a standardized regulatory approach. But I think there's also

Orlando Wood (54:06.872)
Yeah.

Dan Jasnow (54:18.512)
It's not always going to be deceiving to be engaging with an AI agent. I think, right.

Orlando Wood (54:24.642)
Yeah, people will just start to expect it. And if the AI agent laughs at my joke in a human-like way, I should just take the win, you know?

Dan Jasnow (54:33.072)
Yeah, and there are lot of circumstances where, you know...

I think people just want to, they want to move through the technology, they want to move through the process as quickly as possible and in the most efficient way. And you can think about, you know, a delayed flight, a canceled flight. You know, you don't necessarily, sometimes obviously the human interaction is really important and that warmth and et cetera, that experience can be really unique. But in a lot of cases, I'd say the vast majority of cases, you know, you are, you want to get from, you know, from start to finish and you want to do it efficiently and you want to do it in a way that works.

Orlando Wood (55:00.397)
Yeah.

Dan Jasnow (55:08.63)
And so I think if these agents can do that and if they can do it correctly, if they can do it quickly, that's ultimately a good thing for the consumer.

Orlando Wood (55:20.334)
Are there any jurisdictions out there that are getting IP policy meaningfully correct? that you go, that's, you know, obviously the EU was very quick to act with their, you know, AI rules. They're a bit outdated now, but at least they were quick to act. Is speed what's needed or there are some places that are just really actually making sensible decisions and you go Sierra Leone. That's where we need to be looking at. You know, Thailand, whatever it is, who's getting it right?

Dan Jasnow (55:34.756)
Yeah.

Dan Jasnow (55:45.216)
Yeah, yeah.

Dan Jasnow (55:49.965)
Yeah, so...

I will answer that by saying first what I think is where we have gotten it very wrong, which is, and which might be the first AI case that the Supreme Court takes up, addresses the question of whether or not something that is created with AI is subject to copyright protection. The Copyright Office, particularly under the previous librarian, that there can essentially be no

copyright ownership when AI is involved. I think that is, in some cases, that's the right answer, but not in all cases. I mean, if you look at the creative content that's being produced in part with AI and the level of human direction that goes into these tools, just like, you know, it...

Orlando Wood (56:38.958)
Well, that's their litmus test, isn't it? Meaningful human intervention, and then we can discuss what meaningful.

Dan Jasnow (56:45.184)
Yeah, but you there is a case we've, you and I maybe talked about in the past, but the Zario of the Dawn example where a graphic novel where a human artist prompted mid-journey over and over and over again, spent hundreds of hours creating AI-generated images and going from something that was initially very rudimentary to something that was a polished finished product, all using AI but prompting it. And the Copyright Office said, that was created with AI, so that's not subject to copyright protection.

I think that's very clearly the wrong answer. There are gonna be some cases where something generated by AI doesn't have human involvement, the requisite human authorship, and there are gonna be some cases where humans are using AI as their creative tool and imbuing that AI-created product with tremendous amounts of human authorship and human creativity. And so, I think, just like every other copyright analysis, we need to be looking at what was

Orlando Wood (57:36.424)
Mm. Yeah.

Dan Jasnow (57:45.23)
the process, not what was the technology. So, you know, we don't say if you created this photo with a digital camera that, you know, that was, that's not subject to copyright protection. We say, you know, you does it have the requisite levels of creativity? So that, I think, is the wrong approach and the Supreme Court may consider that question relatively soon. Some other jurisdictions, I think, have gotten that right, but that is, you know, something that

Orlando Wood (57:48.824)
Yeah.

Dan Jasnow (58:14.833)
there's been variation internationally. You know on the regulatory piece, the sort of general consumer protection regulation, I am not a fan really of the European approach where it's highly prescriptive, you know, you're trying to create categories of these types of models. I think actually the much better approach is the much more flexible approach which is something like

what we already have in the US in the Federal Trade Commission Act, which is section five of the FTC Act, which says unfair and deceptive acts and practices are unlawful. And that is a great standard that can be applied to any type of technology. It is very fact intensive. You look at the analysis, you say, are consumers being materially deceived or misled? And that applies to AI use cases just as well as it applies to any other type of technology.

you know what we should be focused on in terms of how we regulate these things and with maybe some additional You know adjustments, you know along the margins, you know anything that might be left out there That's not properly sort of caught within that framework But ultimately, you know

Orlando Wood (59:24.814)
Hmm.

Dan Jasnow (59:28.547)
Whether or not we have good policy outcomes depends on who is in charge of those agencies who are responsible for enforcing those types of authorities. So that certainly is not happening today, but it could in the future. And I think properly resourced, that is a better model.

Orlando Wood (59:37.955)
Yeah.

Orlando Wood (59:47.95)
Yeah. Do you have time for one more question? What's coming up? What are you excited about that you're seeing at 11 labs that you're allowed to talk about? Obviously, or if you're not allowed to talk about anything specific, what are the things that you're excited about in the future? Being at a place like 11 labs and seeing the from the front lines of what we're going to see technologically over the next little bit.

Dan Jasnow (59:50.425)
Sure, yeah.

Dan Jasnow (01:00:11.321)
Yeah.

Yeah, I think the most exciting thing that I'm seeing is that I think we have crossed an initial threshold where previously if there was a lot of hesitancy in some of these creative communities among major rights holders, Hollywood, publishers, media and entertainment space, I think there's more of a recognition today that we can be really great partners. And I don't mean just we, like 11 labs, but the AI industry, generative AI, and some of these rights holders.

And I think I think now that we are past that period where you know We've we've sort of built that initial trust and now we're ready to start dancing right we're still ready to start You know exploring what we can actually do together, and I think that's going to lead to two really great opportunities for rights holders and for consumers and if you if you look at it in our case, know where you know one of our Another North Star for us is to make information as widely available as possible and as many

languages as possible. There are tremendous opportunities for things like audiobooks, ebooks, huge volumes of creative works that historically have only been translated into a very few number of languages. Very few percentage of books are ever even converted into audio. And we are able to do today with some of our tools,

conversion, translation at a scale and at a cost that is miniscule compared to what we previously have seen. so, you know, if you, and especially I think if you step out of sort of a US-centric view of this, the ability to share content globally, to leverage existing works, to really leverage those existing IP rights, right? And so, you know, so much of this conversation is about replacement,

Orlando Wood (01:01:58.318)
Mm-hmm.

Orlando Wood (01:02:09.08)
Mm-hmm. Yeah.

Dan Jasnow (01:02:12.045)
really what's exciting here is that you know the market opportunity for these works whether it's film or music or or

books, look at that globally, instantaneous translation, very low or very inexpensive audio production. The opportunity there to share content at scale is just tremendous. And I think we have just begun to really explore the potentials of those partnerships. And I think this year we're going to really start to see that.

Orlando Wood (01:02:47.032)
Yeah, well, that's fantastic. I'm so excited and I'm excited for you and the new gig. I know you're kind of six months in now, but it's you know, it's it's very exciting. I really do think it's a tremendous product and it's it's so rare to, you know, have carved out a lane where you guys are so clearly ahead of any possible competitor. It's very exciting. So yeah, looking forward to seeing the next next tooling come out. I've already got my voice.

professionally in there. I've, I've, I've not gotten as far as sort of putting out a podcast with only my digital voice, but I'm thinking about it. It's not a bad idea. Maybe we should do that.

Dan Jasnow (01:03:17.155)
Good, good.

Dan Jasnow (01:03:21.857)
Yeah, well we could translate this one, we could translate it into any languages you want and get it out there. But thank you for having me Orlando. We should, that's what McConaughey is doing. So yeah, no thank you for the opportunity Orlando, it's been really great talking with you.

Orlando Wood (01:03:35.128)
Fantastic.

Orlando Wood (01:03:38.618)
it's been a real pleasure. Thank you so much. Thanks, Dan.

Dan Jasnow (01:03:41.795)
Thank you.