ElevenLabs CEO/Co-Founder, Mati Staniszewski:The Untold Story of Europe’s Fastest Growing AI Startup

1. 0:00 – 1:20

   Intro

   1. MSMati Staniszewski0:00

      So, we crossed 200 million. (cash register dings)

   2. HSHarry Stebbings0:02

      Whoa. (cash register dings)

   3. MSMati Staniszewski0:03

      20 months to 100. (cash register dings) And then, uh, 10 months to 200. (cash register dings) (baby laughing)

   4. HSHarry Stebbings0:07

      Today, we have one of the fastest-growing AI companies in the world, ElevenLabs. And I'm so thrilled... ... to welcome their co-founder... ... Matti, to the hot seat today.

   5. MSMati Staniszewski0:15

      (cash register dings) Pre-seed was tough. Pre-seed was hard. I think we spoke with a good amount of double-digit investors, between 30 to 50.

   6. HSHarry Stebbings0:22

      How much did you raise in the pre-seed?

   7. MSMati Staniszewski0:23

      We raised $2 million.

   8. HSHarry Stebbings0:24

      Do you remember the price? (cash register dings)

   9. MSMati Staniszewski0:25

      $9 million. For us, even from investor, from product perspective is like, especially now, the speed of execution, speed of investing is- is the only thing you have. Our biggest contract is around $2 million, and they are mostly in a call center, customer support, personal assistant space.

   10. HSHarry Stebbings0:40

       What do you believe that most around you disbelieve?

   11. MSMati Staniszewski0:43

       That you can build a company from Europe at the global scale.

   12. HSHarry Stebbings0:45

       How do you answer the question, "Why won't OpenAI just do this?"

   13. MSMati Staniszewski0:48

       They definitely will, but I think they, you know, they- they lack the... (guitar music plays)

   14. HSHarry Stebbings0:58

       Ready to go? (upbeat music plays) Matti, dude, I cannot wait for this. I've wanted to do this one for a while, so thank you so much for joining me today.

   15. MSMati Staniszewski1:10

       Harry, thanks for having me. After working on a number of projects together, I'm so happy we can finally speak.

   16. HSHarry Stebbings1:14

       Dude, I am so thrilled we could do this. Now, I spoke to so many people beforehand, and I have to say, Luke was phenomenal

2. 1:20 – 6:23

   Early Life & Founding Origins

   1. HSHarry Stebbings1:20

      on your team in prepping me so much. But I do want to start a little bit with the origins pre. You grew up in Poland, and this was Luke's... How did growing up in Poland impact your mindset towards the world and company building first?

   2. MSMati Staniszewski1:33

      You know, in Poland, the... It's a, it's a very different, much smaller world. So it's, um... For- for growing up, I was in a very lucky position where I was in suburbs of Warsaw, then I went to high school in Warsaw, then kind of started seeing more of that world, which, um, which was, of course, incredible because that kind of opens your eyes of like what's possible, that there's so much world beyond what you've seen in, uh, smaller cities. And I think in a similar way as you think about building ElevenLabs now, it's- it's, uh, the scale of what- what we don't know is still ahead of us, and this is, like, what's exciting, that we know that if we climb those- those traditional hills, the other mountains, we will see increasingly more. Um, and I think the second thing was in, um... I saw the transition where in- in suburbs of Warsaw, it's in like a public school, so, uh, school where- where you would have just any- any kids that would live in the area. And then high school and where I met my co-founder, you would have, um, a- a slightly different crowd, like a- a crowd of- of people that would win competitions, uh, need to go through a few steps to get there. And- and suddenly that like kind of density of talent was the most motivating factor to like explore more, learn more and do more, which now we are trying to, of course, replicate at- at ElevenLabs where, and any other company I'm sure too, is like you try to keep that density as high as possible because at the end of the day, that's what's most motivating being part of this incredible set of people.

   3. HSHarry Stebbings2:58

      I was most inspired or given hunger because my dad always said, like, when we went into Chelsea, like, "Oh, it's- it's a better life here (laughs) and like this is where people who've won live." And I didn't live there and I wanted to fucking live there, and it fed me with this like insane hunger. Where did your insane hunger come from in those early days?

   4. MSMati Staniszewski3:19

      The early days, I mean, definitely it was, you know, the combination of family, my older brother, what kind of trail blazed going abroad to study, then motivated us to like, "You need to do the same. This is really helpful." And then the second, it was really this community of people in high school. Meeting my co-founder, um, was extremely smart, meeting some of our close friends, uh, and- and like kind of going through this cycle of motivating each other. It's like, yes, we want to study at the best university. Yes, we want to go deeper and crash this exam. And- and, um, I think this was huge, huge element for- for- for all of us just- just- just- just going after it and knowing that maybe it's possible. Um, at the time, at the time everything felt distant and it's still, you know, many of the things do. But- uh, but I think without that- that kind of motivation through the people, would- would be not possible.

   5. HSHarry Stebbings4:10

      Speaking of the motivation through the people, you have a very special relationship with your co-founder, Piotr?

   6. MSMati Staniszewski4:14

      Yes.

   7. HSHarry Stebbings4:14

      Yeah. So I have to ask then, you come together, how does the idea for ElevenLabs come to be? Was it truly inspired by bad movie dubbing?

   8. MSMati Staniszewski4:24

      There's two things that coincided for us starting ElevenLabs. One was through the years prior, he worked at Google, I worked at Palantir. We would meet for a hack weekend project together. So we would try to explore a new technology. One was in a recommendation space, one w- during the crypto height, we built a crypto risk analyzer, which wasn't very easy and didn't work very well. And then- and then once we did the project in audio, which was the idea was, can it analyze how you speak and gives you tips on how to improve speaking? Um, and that kind of opened our eyes to what's possible in the technology space there. And then fast forward later that year, so we did the hack weekend project in early 2021. In late 2021, the moment hit from Poland, which is, wow, all movies are still dubbed where you have all the voices from the original, whether it's male or female voice narrated with one single character. So you have one voice narrating all the characters in a flat, no emotional way. So it's like a audiobook reading of a movie. It's a, as you can imagine, a terrible experience, and something that we knew combining the experience from working on audio and now knowing this is a problem that, okay, this will change. A few years from now, as you- as you- as you imagined that kind of advancement, all the voices will have original emotions, intonation, sound incredible. And that kind of kickstarted this idea and of course then it expanded to, okay, we need to fix the research layer to actually make it happen. Um, then we started with a lot of creative platform work around ElevenLabs, and then expand it to agentic platform work of how now the interaction is shifting where voice is this big interface for the technology around us. So-Well, you know, it was very dubbing-specific originally, it kind of expanded through to everything voice today.

   9. HSHarry Stebbings6:09

      I mean, it must be the most boring movie experience, having one single voice for everything.

   10. MSMati Staniszewski6:14

       It's terrible. It's ter-

   11. HSHarry Stebbings6:15

       So, what do you do? When you land on the idea, then, in terms of your next steps, you move into the research layer and determine whether it's possible to actually do this?

3. 6:23 – 17:00

   Building the Product & Early Challenges

   1. HSHarry Stebbings6:23

   2. MSMati Staniszewski6:23

      The way we've approached this was we, um, we first tried to, like, do two things in parallel. My co-founder was, "Okay, can we use existing technology, stitch it together, and create a dub of a movie in a better way?" And it quickly transpired that you get a good effect, but it wasn't brilliant. So, to actually fix it, let's take a step back and fix o- one of those components and make it great. At the same time, when he was doing research, my mission was, does anybody actually want a dubbing product? So, I was emailing other YouTubers, uh, uh, um, getting the, the, the emails, scraping them and trying to message them, "Hey, if we had a dubbing product to make your movies available in all languages, would you be interested in that?" And, um, there was a, not t- there was like a, a, a roughly, it was like 15% reply rate initially from the, the, the first batches that we sent. All were personalized. Um, I mean, we sent thousands of them. But the, the interest was, like, la, lack-, lackluster. All of them were just like, "Oh, I don't fully believe this is possible. Can you send us a sample? It would be great, but how will I operationalize this? YouTube doesn't support it." So, it's like, it was somewhat of an excitement, but not like a burning problem, I need to solve it. But then the second thing happened. So, we started sending the samples. Uh, we started speaking more with the, the, the YouTubers and it quickly came that what they actually want help with is much simpler. They want to post-produce and correct if somebody said something incorrectly. They, they want to understand how the script will sound before they need to produce it. They might not want to speak at all, and they want to voice over, over the movie. So, a super simple problem which didn't include anything with language changing. And, um, and of course my co-founder and us, he, he dived into the, the, the research, real- realized, and he's an incredibly smart researcher, that you can actually buil- build completely new text-to-speech model which will be a lot more emotional, intonative that, um, that will make that narration a lot better. So, let's put the dubbing problem aside, and as we see, there's this huge other problem with, with a lot of the creators. Let's solve this problem first and bring that research just on the text-to-speech layer. And then, you know, kind of, as usual, as we dive deeper-

   3. HSHarry Stebbings8:33

      And that's only possible by creating your own models.

   4. MSMati Staniszewski8:36

      There was... We... Yeah. Quick answer is yes.

   5. HSHarry Stebbings8:38

      Mm-hmm.

   6. MSMati Staniszewski8:38

      The, all the models at the time just, you, you could tell immediately it's like uncanny valley, not very good. You cannot replicate voices.

   7. HSHarry Stebbings8:45

      Dude, how should I know as an investor to say... Again, I use this show just to get better as an investor (laughs) . How should I know as an investor whether a problem needs its own model or whether you can just leverage alternative existing models?

   8. MSMati Staniszewski8:57

      Well, the... I think the answer will definitely change now to, like, 2022. So, that was the year where nobody e- yet, like, kind of was thinking about AI. This was still kind of downfall of Metaverse, crypto days. So, you were, you were still... The attention to, to the models was w- there was, was like n- none in the public eye. You know, ChatGPT happened l- beginning of 2023. That's when it kind of all started spiking across. So, at the time, um, at the time, you didn't really have a choice. Like, you, you, you c- you could, you, you, uh, you could kind of n- ... You knew, um, as a user, as an investor that what existed out in the market just wasn't very good. Um, and then there was more of a question, will this team be able to solve and create something better? Uh-

   9. HSHarry Stebbings9:39

      If you were creating ElevenLabs today, what would you do architecturally differently given the architecture that we have today?

   10. MSMati Staniszewski9:46

       I think still a current question is to what extent you, uh, you continue on that kind of single modality space, where at the time it was mostly you train a dedicated model for speech or dedicated model for image, video, et cetera, uh, over what is now more of a theme where you kind of train a more of a multimodal approach, where you combine the reasoning and speech together to create even better speech experience. So, that's n- our most recent generation is effectively that, the Eleven V3. If we, if we had that then, I think we would have probably skipped few steps and our experience would have been even, even better.

   11. HSHarry Stebbings10:21

       S- so off-tangent, but I'm just concerned about bluntly the plateauing progression of models. And, you know, GPT-5 was the, the embodiment of that. You move from, like, incredible feature descriptions to a description on cost. And when you have a description on cost and efficiency, it's like, "Oh, we've, we've hit that."

   12. MSMati Staniszewski10:37

       Right.

   13. HSHarry Stebbings10:38

       Do you agree? And are we in an element of incrementalism and plateauing?

   14. MSMati Staniszewski10:42

       In voice, I think the, the ... So it's, you know, there's a little bit of, like, it depends. It depends on the use case. But to make it more concrete, there are use cases like narration. In narration, we think it's plateauing. You... The, the new model generations will not make narration of an audiobook drastically different. It will still be, it will be still, um, in a similar quality. But I think in general, the point is right, where if you just do research, eventually it will commoditize and eventually that advantage you can deliver from research isn't enough. So, you really need to build product. And as you think about Eleven, we combine both together.

   15. HSHarry Stebbings11:17

       Do you think we're at a stage where scaling laws no longer equal, uh, an equivalent level of progression?

   16. MSMati Staniszewski11:23

       Well, it's in a, in a, in a, in a, in a biased way, I feel like the progression is, is, is, you know, still, like, just scratching the surface where it's, it's just, just the moment of, like, still, um, seeing that kind of escape curve of, of, of AI getting adopted everywhere is just getting started, while scaling laws continue in the same way. So, I do think there are-

   17. HSHarry Stebbings11:43

       But there's a difference there between adoption and development. So, I agree with you there in terms of the adoption, but actually the development progression.

   18. MSMati Staniszewski11:52

       Got it. I think, you know, it's probably different answer on LLM space and different answer in the, in the voice space. I think in voice space, you are still, you, you are still in, like, not ... Y- y- it's still not a slower rate. Um, and then there's an interesting variation of, like, as they all combine-What does m- th- that mean for the world? Like, a slightly deeper understanding of everything around you? Um, but I think in voice you still will see pretty, pretty, pretty quick curve. On the lumps, I agree, it's probably a little bit flattening to- to some extent.

   19. HSHarry Stebbings12:24

       I spoke to Andrew Reid before the show. I had a great chat to him last night and he said, um, "Ask him about, like, why can't OpenAI do it?" And I remember chatting to Ciaran about this. Ciaran obviously being my partner, uh, and there were two reasons why, you know, we didn't work together. One is the small check which s- he very kindly offered, um, and the other was, like, no offense, it was super fricking early, and it was like, well, why wouldn't OpenAI do this that, no offense, two very young guys in London are doing? Like, how is this not gonna be done by them? How do you answer the question, "Well, duh, why wouldn't OpenAI just do this?"

   20. MSMati Staniszewski12:58

       They definitely will do- do something, but I think they, you know, they- they lack the- the- the genius that, uh, that- that- that- that- that I'm happy my co-founder and the team has. But I think the true- true- true longer answer is, one is focus. You really, y- I think in the early days especially, you know, when we spoke then, there was just so many different things you could do in AI space. And I think we took this bet that we want to really own and win in the voice AI research and product space. All our- our work is- is- is directly tied to- to- to- to voice. Then the second thing is, I actually think the number of researchers in the world working on voice and being exceptional is super small, probably like 50 to 100 people at, like, this top level. And, um, and- and Piotr and, um, and- and- and people around him were able... So Piotr was able to assemble one of the best teams in the space. I think we have five to ten people that are in the top- in the top 100. Like, they're... It's like a mighty team, which- which I'm kind of time and time again surprised how incredible they are, uh, and to make it specific, I mean, text-to-speech was one. They kind of have blown everything out of the water. Then speech-to-text is beating OpenAI, Gemini on benchmarks. Now music, which is again, n- something that no big company has- has yet been able to crack or do. Um, so it's a- it's a- it's incredibly mighty team. Um, and then even if, I think if OpenAI does, you know, actually do something on research, and I think they are trying, then that product layer is where- where you really need that- that big advantage of. If you are in a creative space, if you do narration, if you do voiceover, if you're Dub, you go through so many additional steps to really make it perfect. We do that well. If you're building a voice agent or conversational agent, you need to bring knowledge base, integrations, functions. You need to then deploy, test, evaluate, monitor. All of these pieces are coming together in a platform and I think OpenAI is, you know, not- not investing as much time. They could, but they, but they, but they, but they aren't. So combination of mighty research team, speed of execution, and then actually focusing on- on the use case in a- in a... I think in a deeper way.

   21. HSHarry Stebbings15:04

       You mentioned that kind of the mighty team. Those people are very valuable people, and an OpenAI or an Anthropic or a Meta would pay a lot of money, it would seem, for people like that. Do you worry about the war for talent, and how do you think about retaining talent when there's hundreds of millions of dollars going across the table for them?

   22. MSMati Staniszewski15:24

       I think the talent, especially on the research side, is- is- is- is- is- is incr- the- the impact that those people can create is- is- is out of- out of this scale across any of those companies, especially in the early days. Of course, to some extent as you, as I think about Meta and others, they are of course paying for the know-how as much as they are paying for the raw talent too where getting those early people gets you, um, some insight into the models and architecture that you can then bring across and accelerate. So in early days it's- it's- it's more valuable than late. Um, but, uh, but I think in our case, the- the reason I think it's- it's still valuable is one, the upside for Eleven Labs just getting started. So I hope we'll be able to compete with any of those companies in the- in the future at- at- at- at the same scale. Two is, um, what you- you do have which- which I think is important as I think about any of those companies is, like, how close you are from developing research to actually deploying research. And at Eleven Labs, if you are actually creating new models, they get to production and are some of the most important things that our products work on almost immediately. So the gap is super quick. Um, well, and, you know, bigger companies who go through their usual corporate re- red tape to- to get any of that- that work done. And then three, um, I think we- we- we now have, like, a- a very small and mighty team that can learn from each other and- and- and move pretty quickly. I don't think there is a guarantee you'll get that in- in some of the other companies. I think they are optimizing for a lot of people, but not necessarily the right people.

   23. HSHarry Stebbings16:48

       We're gonna go to team size later. I do want to stick with some chronology. I have so many questions that I want to ask you, but some chronologies. We mentioned that kind of the V1 and building to beta, and I think it was, uh, where

4. 17:00 – 30:20

   Pre-seed Fundraising

   1. HSHarry Stebbings17:00

      was it? Uh, 2023 that you raised the pre-seed?

   2. MSMati Staniszewski17:03

      Mm-hmm.

   3. HSHarry Stebbings17:03

      January '23. Um, how did the fundraise process go? You're two young guys saying, "Hey, we're gonna do this." How did the pre-seed fundraise go?

   4. MSMati Staniszewski17:13

      Pre-seed was tough. Pre-seed was hard. We... It was, you know, similar question that you asked. It's like, how will we fix research? The second question, which was very interesting, we think the market is very small for what you are solving. Which at the time was like, you know, it's like a combination of AI voice, like nobody was thinking about that. So- so that was, that was, that was surprising to us, which we were very disagreeable. And third was defensibility. Will it actually be better than what incumbents from the- the FA&G companies will actually solve or- or like how will you out-compete it on long term? So these were the three. So it was tough. I think we spoke with- with a good amount of- of double digit, um, investors and, uh...

   5. HSHarry Stebbings17:56

      Double digits of investors, and they said no?

   6. MSMati Staniszewski17:59

      Yeah, there was like f- f- w- with between 30 to 50 and it was double- double- double hard at the time because we- we... Early 2022, we got an offer from one of the accelerators in US, not YC, and, uh, and we are thinking like, should we break, take it or not? And we decided no, we are rejecting. We think we can create something- something- something, um, more valuable and we don't need that help at the time.... so decided to go independent. And, uh, this triggered this more stressful time where suddenly now you need to raise money. We started sp- spending a little bit more on GPUs. We hired our first few people. It's like all of that was going from our savings from... which we are lucky to have from, from Google and Palantir. But also like, okay, now it's, now we're, it's actually getting a little bit more risky and we want to double down. We want to invest even more, so we need money. And, um, and yeah, it was, uh, it was, uh-

   7. HSHarry Stebbings18:46

      How much did you-

   8. MSMati Staniszewski18:47

      ... a tough one.

   9. HSHarry Stebbings18:47

      ... raise in the pre-seed?

   10. MSMati Staniszewski18:48

       We raised two million.

   11. HSHarry Stebbings18:49

       Two million. Do you remember the price?

   12. MSMati Staniszewski18:51

       Nine million. It was-

   13. HSHarry Stebbings18:52

       Post. Nine million post.

   14. MSMati Staniszewski18:53

       Nine million post. The end up, the amount was exactly 11% of the, the eq- the, the equity that, that, that the first investor was buying, and then the other ones were layered in.

   15. HSHarry Stebbings19:05

       Oh.

   16. MSMati Staniszewski19:06

       So it was like just over a million. It was, uh, it was a, a-

   17. HSHarry Stebbings19:09

       And this was 2023?

   18. MSMati Staniszewski19:11

       That was 2022.

   19. HSHarry Stebbings19:12

       This was not... Yeah, but dude-

   20. MSMati Staniszewski19:14

       No, the-

   21. HSHarry Stebbings19:14

       Dude, I'm so used to like stories like this on the show where it's like HubSpot and you're like, "When was it? Oh, 2006," or 2008 or '09 or whatever. But this was fucking recently.

   22. MSMati Staniszewski19:23

       Oh, yeah. I, yeah. It was only, only four years ago which, which, uh, I mean, it was so... You know, in the end we found... We had Credo Concept, co- lead co- concept here, here in UK. Credo from ZE. We had, um, one of, of friends, um, of Piotr's from Oxford, Peter Chaban, who invested, was a co-founder of, of Polkadot which is a cryptocurrency. So we had like a good set of early believers. But to get there it took us, it took us a bit, a bit, a bit longer.

   23. HSHarry Stebbings19:49

       So you raised the two million then. What happens then? You start building out a team, you start investing in GPUs.

   24. MSMati Staniszewski19:55

       Exactly.

   25. HSHarry Stebbings19:55

       Just take me through that.

   26. MSMati Staniszewski19:56

       The, the, the main reas- the main reason we raised was to accelerate what we would, what we want to do. First one was building a small data center, so we built one. And back then it was, I think, in Poland that we've, we've invested few, and then quickly after moved and started b- buying some in US. Uh, and then team. We started hiring, not many but it felt like a lot at the time. It was like additional two people, I think. (laughs)

   27. HSHarry Stebbings20:20

       (laughs) Two people. That's quite a lot.

   28. MSMati Staniszewski20:22

       (laughs) Yeah.

   29. HSHarry Stebbings20:22

       Well, don't go, don't go too far there, Matei. Um, did you... 'Cause w- you kind of did the pre-seed raise and the beta launch at the same time. Did you have product market fit on the beta launch? Were there early signs that it was working?

   30. MSMati Staniszewski20:35

       Oh, yeah. That's, uh... Okay, so got, got... The, the good, um, good timeline, uh, piece there where we actually raised the round in mid... Like a Q3 of 2022, but we announced it in Q1 of 2023. And the reason we do that, and we do that all the time, is that we don't want to announce a round just for the sake of announcing a round. Our philosophy was always the round should have another purpose which is bring the product out and, and help you get the product into the users. Um, celebrate a set of customers to show that you are arrived in a specific sector. Uh, bring a new research model into play. So, like, every round that we would do would always tie into a product announcement. So we actually hold in, hold on for announcement for a few months until we had our beta release, and then we triggered which was January 2023. Um, but your, to your question, initially when we worked on dubbing and, uh, and that first early days and we're like emailing, we didn't have product, product market fit. It was like very clear that... You know, people were slow to reply. Then we sent samples. They weren't engaging, so we, so we didn't through the 2022 period. And then when we shifted from dubbing into narration, voiceovers, then the product market signal started to hit. "Maybe it's there." We, we had a first, um... And I remember distinctly. There was three things that happened. First, we did a blog post around the first AI that can laugh, and we sent samples that got picked up by newsletters and people were like, "Wow, this is incredible." And the next day we had like thousand people on our waiting list. Then the second thing happened, we invited the first, uh, 100 or so users to, to ElevenLabs to test it out, and we had this audiobook author who, uh, who joined the platform. Our platform is like this small text box where you can type in text, tweet length, and narrate it. And he would copy-paste his entire book 500 times, download it, stitch it together, then he released it. He released it on the platform at the time AI was banned. It passed through as a human content and, um, and then it started getting reviews that it's great, and then he came back saying, "I want my other, uh, uh, uh, book author friends to do it too." So it's like, "Okay, we are onto something." And then, uh, and then we launched in January 2023 publicly. We knew... We started getting more of that signal that creators, narrators love, love the work. And to be honest, I think from that moment on- onwards we s- saw, like, a clear momentum. And then there was like, you know, more events after the media picked it up, then more creators picked it up. By the moment of release we knew that it's valuable. Uh, and then maybe last thing on that, it's like the, the product market fit concept was... You know, it's, um... Fo- for us we never like fully, I think, know what this actually means where... We, we knew users loved our product, uh, at the time but, uh, but I wouldn't call it at the time we had product market fit as we were thinking of like, "Okay, how do we make sure it's providing value for the next 5 to 10 years? Then we'll decide..." Th- this will be like something where we would call, "Okay, this is clear product market fit. We think this is like self-sustaining for as long as we, as we go into the future." So we wouldn't go by that definition. I think we are now closer to that, but we are still, like, you know, still, still know we can create so much more value.
5. 30:20 – 41:14

   $90M Raise & Working with A16z

   1. HSHarry Stebbings30:20

      And then in June '23, you raised like $19 million from Brian Kim and Andreessen and then Nat and Daniel from, um, the... I can't remember what it's called. NDFG or something.

   2. MSMati Staniszewski30:31

      Uh, yeah, NFDG. It's, uh, Nat Friedman and Daniel Gross.

   3. HSHarry Stebbings30:34

      There you go.

   4. MSMati Staniszewski30:35

      From AI

   5. HSHarry Stebbings30:35

      NFDG. Wonderful. Brilliant name. Um, and...That was quite a shock because it wasn't a London fund, it wasn't an index, it wasn't Excel. How did that round come to be?

   6. MSMati Staniszewski30:46

      Yeah. The NFDG is, is now, now I feel like it's a good name, but I was pitching out on changing that to like a different name for a while. Maybe now it's not possible anymore, but, um, the, the... So it was like the interest started around kind of March 2023. We had a lot of the investors that were in close conversation in the past, like re-approached us. We effectively would, would, um, would, would, would be a little bit more in the waiting mode. Like let's see how we, we, we get through and let's optimize for the partners that we truly want. And what we truly wanted was a combination of making sure that people in globally in SF trust and know that we have arrived, like we are a company that can be trusted, is building something ambitious. Um, and then, of course, people that we would admire to, to be able to work with that created something special. And that was definitely in that, in that, in that category, in the latter category. And it was a very fun thing because he approached us. We previously-

   7. HSHarry Stebbings31:45

      Does he just DM you? Like how does he just approach you?

   8. MSMati Staniszewski31:48

      He emailed us. He emailed us, but then I, I met him here in London. He flew in, and, um, it was like a weird setup where he emailed, he emailed me, but it wasn't clear whether it will happen and it happened another last time. And then I see him in London in one of the hotels where he was staying, and the first thing he was saying was, uh, which is the only investor ever to, to, to do that was, "So I tested your APIs and this thing doesn't work, this works. Voice here wasn't good. Stability wasn't very clear parameter." So he's the only person that tested our APIs, decided that it was actually valuable what we are doing, gave me feedback on like what we should change, and then told me like, "And I'm keen to invest." Uh, which, which, uh, which, of course, triggered a conversation of what's the, what's the, what's the vision, what we are after. And what's special about Nat is that you can tell him very little and he kind of has this b- beautiful ability of, of infilling all the context around it, comparing that to any of the past experiences and kind of bringing any of those experiences that then might be relevant to you. But he gets it immediately. So he was, from the first conversation, he knew kind of exactly what we are after and challenged us with the right questions from the start, and then he moved very quickly. He has very keen, um, to, to partner with us. We are keen to partner with him. We are trying not to show it too much, but, but we were. And then, at the same time, we know that Nat was, uh, Nat and Anil were, were more, more perceived as the angels while we admired them from their GitHub and Summer and days, they, th- that was like still like an angel category for many of the externals. So we knew this is our chance to bring a huge investor to actually position us as a top company, and that's, that's... We, we spoke with all the classic tier, tier one funds, and a16z was just one of the most passionate-

   9. HSHarry Stebbings33:36

      So you spoke to all the funds in London?

   10. MSMati Staniszewski33:37

       We spoke with few funds in London, yes.

   11. HSHarry Stebbings33:39

       And then Brian flew to London?

   12. MSMati Staniszewski33:41

       And Brian flew to London. So Brian... I mean, a16z was phenomenal. They, they, they did exactly what we spoke about, which was they've shown us that they cared before they invested. They introduced us to, to incredible people, to some celebrities to work on the voice licensing. So they were like on it for two weeks prior or like week prior, um, already to, to, to help out in any way they could. And then Brian flew to London, uh, uh, uh, and met, met with us and a- and then we signed a, a preliminary term sheet. It was interesting. He flew in, we were negotiating. Also, my phone called to, to Piotr, to co-founders like, "Hey, what about this terms? Can we accept this?" And, uh, and Piotr would, would, would give guidance on what we... where is the hard line, what we can do, and then, uh, and then on speaker we would sign together and, and, um, and finalize.

   13. HSHarry Stebbings34:30

       Do you think speed really is a differentiator as an investor in winning?

   14. MSMati Staniszewski34:34

       Oh yeah. Speed is like for us, even from investor, from product perspective, it's like you need to... Especially now, the speed of execution, speed of investing is, is the only, only thing you have.

   15. HSHarry Stebbings34:43

       This is what pisses me off. I'm happy in that world. I can move supremely fast and invest a lot of money. But what I find more and more with founders is they want to run a road show and they want to do a 10-day process meeting investors and then have another week to decide which term sheets they want to take. I'll- I'm willing to give you a term sheet tonight, save 17 days, but it's more and more common that founders actually want to run road shows. How do you think about that?

   16. MSMati Staniszewski35:11

       Yeah, this, this... Um, now in a lot of, a lot of great companies get built on Eleven Labs, so I get to be in the conversations now more frequently where they would raise money from our investors. Uh, like recently we had a, a company raise a series A that has built a healthcare voice bot, which would effectively help patients calm them through their conversations in a, in a, in a incredible, incredible way. Then there's like another company building for, uh, healthcare customer support. So quite a few. And now I'm in the conversations of them raising the money, which is an interesting one. And, um, and then usually it's, it's clear that, uh, so- some of, some of the people don't know fully what they are actually optimizing for, and it's good to, to, to talk through. Are they optimizing for the valuation? Are they optimizing for dilution? Are they optimizing for the right brand name to help them out? Uh, is it the network of people that this brand name has? Is it the specific partner? Is it something, something different? And I think it's like pretty distributed. So when I speak usually with founders or our investors ask me to speak with the founders, it's usually that question, like, "What is it actually that you, that you care about for the thing? What do you miss?" And then it's, then it's ............................ pretty quickly, whether, you know, they are going after just the road show. Is there to bump it up because they think the value that they- the term sheet they got is just not valuable? Or is it that they actually want someone else, but they don't want to... they want to keep the optionality of having the first one while, while running the process? Um, but you know, all the, all the traditional simple things.

   17. HSHarry Stebbings36:36

       I'm a romantic. I hate this idea of a road show where it's like, "I'm gonna meet everyone and then I'm gonna compare term sheets and then I'm gonna decide." It's like, what happened to the partnership?

   18. MSMati Staniszewski36:44

       (laughs) The... Well, but it's, you know, the truth is, the other way is, is true, where a lot of the investors, um, want to give the term sheet, but that is before any show of partnership. So, how do you know that this is a good partner? How do you know whether they are giving you good terms? Especially our first-time founder, you never run this. You don't know how much you are valued. You need to have some relative comparison, and this can happen from conversations with other companies, other founders, or, um, comparing yourself to other companies in the field, or running, running some form of a process. Roadshow probably sounds pretty bad because it sounds like a lot. You probably... Like, the way we approached fundraising was always, you know, especially in the early days, like, queue few investors we really care about, and first speak with maybe one or two that maybe aren't a priority so we understand whether, like, w- we are at all kind of perceived in the, the, the, the kind of the, the messaging that we want to put across makes sense. If it does, move straight to the, the, the top ones.

   19. HSHarry Stebbings37:42

       How do you feel about exploding term sheets?

   20. MSMati Staniszewski37:46

       I, I don't like them, but we got a few. We had a few who-

   21. HSHarry Stebbings37:49

       Do you understand them?

   22. MSMati Staniszewski37:50

       I do. I think I do. I, at the time, I didn't, but, uh, but I do. It's like, I mean, I don't like them still, but, but we... I won't name the people, but, uh, uh, but so one example was a, let's say, a smaller fund. And they, they, they, they felt that if they get as a term sheet, we'll use that to bump other term sheets.

   23. HSHarry Stebbings38:14

       Very often happens.

   24. MSMati Staniszewski38:16

       And that does happen. Um, so I got it. It was like, you know... And, and they came from the approach of like, if it goes too high, they won't be able to invest, and they didn't want to be part of that. Um, so that happened. And of course, you know, it's good pressure.

   25. HSHarry Stebbings38:28

       Very often being an investor, being the first term sheet's the worst because you're used as the stalking horse where they then go, "Hi, I've got a term sheet, I've got a term sheet from a great fund." They never name the fund, and it's... And, and you get no credit for being the first, and then you're just trampled on by everyone else that comes on.

   26. MSMati Staniszewski38:44

       Oh, yeah, that's what I... That's what I... I don't like founders that are trying to do that too much, like that if they are trying to get you to... get the first-term sheet to bump the other ones, I think it's a wrong approach. I think you should... Like, it's, you know, okay if the... if it's like order of magnitude off, it's probably wrong, or it's like, you know, 5X, then yes. But if it's like plus/minus 20%, that's like a wrong optimization. There's just so much more value that the investors, partners can give than the, the money itself that I think should be optimized for.

   27. HSHarry Stebbings39:08

       You've got Andreessen introducing you to celebrities, and you've got Nat and Daniel, two of the brightest minds in AI, who are querying APIs that no one else does. Are the Americans just playing a different game to the European VCs?

   28. MSMati Staniszewski39:22

       Hm. We love our partners. I mean, we have... A- Andreessen team is so, so incredible, their network. I mean, now, Sequoia with Andrew is just another, another level two, iconic team. Uh, also crazy help. Um, NEA2 is just like bringing all the time.

   29. HSHarry Stebbings39:39

       But they're all US.

   30. MSMati Staniszewski39:40

       They're all US. So, I think they are (laughs) playing a different game. I think the... You know, they're... They are all, from our experience so far, a lot more keen to take the risk. Like, you know, even our conversations are always like, "How do we bet bigger rather than, like, optimize for, for, for, for, for the, the negatives?" Um, the one check I always try to do with, with Piotr for any of the partners that we bring beyond the Capital Network brand is, is always for the person you, you partner with or the person you pa- partner with, how will they behave if it's not going well? So, we usually try to get any of the back references for how, um, how did they behave when the company was, was failing. And, um, and all of the ones that we had are, are, are in a very positive. Like, they actually were, were really on the, on the side of the founder, helpful when it was not going well, even more so than when it was going well. And I was doing the same check with some of the checks with, with, with other companies, and, um, maybe this is getting used to that, like, you know, when you ask the kind of the set of companies that went up and down is so high. But in some of the investors we, we spoke in Europe, that check didn't yield the good results.
6. 41:14 – 1:02:20

   Team, Culture & Scaling

   1. MSMati Staniszewski41:14

   2. HSHarry Stebbings41:14

      Okay, $19 million, doosh, in the account. Most people suddenly then go on a hiring spree, and not two people hiring spree, but like a proper hiring spree.

   3. MSMati Staniszewski41:24

      (laughs)

   4. HSHarry Stebbings41:24

      But you favor very small teams. You've said this to me before, but Luke said it to me. Why do you favor really small teams?

   5. MSMati Staniszewski41:32

      I think the first piece is that more people frequently doesn't fix the problem. It's like you, you, you, you, you don't need that many people to do something special. That's the first thing. Second is, you... by keeping... And as an organization today, we are... you know, we have 250 people, but really, it's more like 20 teams, uh, doing go to market of like five to ten people that are just executing on specific project, where, um, where they have higher ownership, they can move extremely quickly, they, um, they see the result and iterate with reality a lot quicker to improve that. And I think this just works in such a beautiful, beautiful way. Of course, there are other challenges with this, but it's great.

   6. HSHarry Stebbings42:08

      Are those 20 teams organized by function or by project?

   7. MSMati Staniszewski42:13

      Uh, a little bit depends. The, the, the... In the product, by like a product, um, product area. So, we have a team working on our studio interface, team that will be responsible for all the core experience when you log in, team responsible for the entire voice agent suite. And within that, there's a team working on some of the more enterprise components, another team working on some of the self-serve elements. So, all of the teams will be organized there on the product area. And then, the other parts, we try to shatter pretty quickly. So, you know, and, uh, we will have a separate team for talent, uh, of course, and a, a team for people. And they will have a high degree of, of independence when, when they actually execute, which, which helps.

   8. HSHarry Stebbings42:54

      You mentioned 250 there. It- it's- it's small given the size of company in many ways, but it's also still 250 people. When was the company culture the worst, and what did you learn from that?

   9. MSMati Staniszewski43:04

      You know, there was, like, moments where, where, where, where I could feel this kind of tension between the go-to-market research engineering, um... And it's like a specific story where... So, I mean, it's a super lucky position where we develop our models and bring them to the customers, which, which of course makes it very special. We frequently will show for the very first time to the world something that was n- not possible and is possible now. In 2023, late 2023, so we did a text-to-speech, we then did voices, so we could recreate a voice. Then we, um, then we create like a basic way for speech-to-text internally. So, we had all the components that we originally were speaking about to create dubbing. And we, uh, we haven't done dubbing yet, publicly, and we gave all the components to some of our customers, so they could use text-to-speech, the voices. And, um, there was, like, a lot of, you know, pressure that, like, it's a valuable technology we can give to, to our enterprise clients. And, um, one of the enterprise clients, we told them that we are planning to launch our dubbing solution combining those components later, later in that... It was, uh, it was I think September, lat- later in that, that month. And, um, and they took the components and released dubbing two weeks before we did. And that was one of, like, the low moments for me, for my co-founder, for the entire team, because dubbing was our story. We told that to everyone. We knew that this is the thing we want to solve first. We had all the components to be able to do it. We're just waiting to optimize it to make it perfect. And then suddenly, this, this, this, this, this, um, partner released it and got all the attention. So, all the, uh, all the, the media, Twitter, all the users were like, "Wow, this is the most incredible thing that has happened. How amazing that you can speak in another language and still sound the same." And you could feel the morale in the company was, was, was low. People were like w- you know, we spoke about this. We knew for, like, almost two years that this is something we want to solve, and how did it happen that the customer had this and solved this first? So, research engineering wasn't happy, like, how did they do it before with the components we gave them? The go-to-market wasn't happy because, like, how did we sell this to the customer? Now we are like all the potential clients we could have had isn't, isn't good. And of course, me and Piotr were like, "Hey, this was our idea. Like, why, why would you do any of the partnerships if all of that is, like, kind of given to, to, to, to, to another, another company?"

   10. HSHarry Stebbings45:36

       Who was the partner?

   11. MSMati Staniszewski45:38

       I don't know if I can, I can mention. And to, to their cred, I don't think the time, like, that we told them it was actually a factor. I think they were just trying to do something quick and... And, like, we, we had a lot of calls with them those, those days, because we were thinking, like, "What do we do? Is the contract give us flexibility to, to not continue?" So, we spoke through with them of, like, "Hey, why did you launch?" St- I don't think it was dictated by our timeline, but the timeline was very close to each other. And they told us that they thought it's a good hack weekend idea, so they gave it to their intern, and they... The intern has built that project, and then, um, and then it exploded. But it was... It di- it did give them, like, in tens of millions of revenue over, over that period of time. So, it was significant.

   12. HSHarry Stebbings46:19

       Fuckers. Uh, what's your biggest piece of advice to a founder who has a moment like that, where you just feel the air come out of the company? How do you inflate the company again after such a damaging blow?

   13. MSMati Staniszewski46:34

       Y- I don't think the first reaction should be that, like, "Hey, everything is fine." I think you should, you should, like, be authentic and tell the team what you are feeling. Like, "How has this happened?" And actually, the, the, the go through, like, "What have we done wrong?" And, you know, that, that frustration was clear. I don't think we... W- I, I, like, I think it was valuable for us to talk through, like, "We are angry at ourselves. This is wrong." And then as we moved from that stage, like, "Yes, we are angry. What are we actually going to do about this?" And, um, and moved to that second stage very quickly. Uh, but I think, you know, I, I, I think I've done a mistake sometimes of just going to the second stage, like, kind of, "It's not a problem, let's just go through and solve it." I think it's actually super valuable to, to, to, to talk through what, what has happened. Uh, but then you knew you need to, you need to go into the second thing, and then, and then in the long term, that will work out, you know? It's, it's, it's then you can... The relentless execution will prove itself out and, and, and show to the world. And the, the worst, if you repeat the mistake, then someone needs to f- feel the, you know, feel, feel the, um, repercussions. But if you've learned from that and haven't repeated the mistake, it's fine.

   14. HSHarry Stebbings47:42

       Given the commoditization of a lot of technology stage, do you feel speed of execution is the core differentiator between those that will win versus those that won't? Or is it quality of research, access to GPUs? How do you think about that?

   15. MSMati Staniszewski47:56

       I think it's both. I think it's both. I think it's... You know, uh, the, the way we approached this in the company is we, we have researched product in the ecosystem that we built, um, which is a combination of distribution and brand. But research for us is a head start. We are investing in research. We'll continue investing in research. We want to be the best across the voice technologies. But all this gives to us is advantage over the competition for the next year, two, maybe three, to be able to create-

   16. HSHarry Stebbings48:23

       How mu- how far ahead are you compared to the competition, do you think?

   17. MSMati Staniszewski48:26

       I think it depends on the use case, about 6 to 12 months, I would say. It's depending on the, on the, on the, on the space that we are in.

   18. HSHarry Stebbings48:32

       Do you think that's a lot or not?

   19. MSMati Staniszewski48:35

       I think that's a lot. I think that gives us... So, that research piece of 6 to 12 months is enough for us to then do the second thing well, which we do in Pile from the start, is build phenomenal product experience.

   20. HSHarry Stebbings48:44

       What percent of your revenue do you think you spend on compute?

   21. MSMati Staniszewski48:47

       So, we've built our own, um, data centers for, for training, and then for inference, we, we of course, um, use some of our, our great partners across, um, across the traditional kind of things.

   22. HSHarry Stebbings48:58

       Why build your own data centers? Most people would just use a CoolWeave or use NVIDIA or whatever that is. Why build your own data centers?

   23. MSMati Staniszewski49:05

       So we did the, you know, we did the maths. In our case it was, if we assume we continue training the models that, the way we want to, so very continuously a lot of those models, um, then second for the data transfers and as we think about continuously bringing more data, um, we- we- we will likely on a two-year horizon break even for having our own and- and- and not renting. Um, assuming, of course, you know, you need to assume some improvement in the GPO infrastructure, but assuming that wasn't too high of an improvement, we would have, we would have been, um, successful, and we were. I think it, uh, it kind of the arrow Y made sense there. And- and now it pays dividends because we can just do more experiments quicker, and, um, you know, this can still be wrong. At some point there might be innovation that kind of breaks that equation, but for- for- for the current time it just makes more sense, more control.

   24. HSHarry Stebbings49:56

       A lot of investors are talking about the poor unit economics of a lot of companies that we see today, whether it's your Replites or your LovaBulls or any of the companies like this, well, any kind of application AI companies, period, really. They generally have pretty tough unit economics, generally speaking. Um, do you think that is a fair criticism or do you think it is simply short-sighted given the changes we will likely see in the cost structure?

   25. MSMati Staniszewski50:20

       Of course, the unit economics in most of those cases that you mentioned is pretty- pretty poor, but I think their strategy is that, yes, one, the models will optimize in cost and then, two, they will be the brands that customers trust and then they can actually use a lot of the signal back. And I think to Eleven Labs' example, too, we, I think our unit economics is- is much healthier than most of those companies. We control our research, our product, and distribution, but it's still, if we had a new model that we need to release ourselves, we would optimize less for- for the cost structure so we can get that metric out quicker than other competitions even can think about creating their own model.

   26. HSHarry Stebbings50:57

       But to be clear, you think the concerns around, um, margin are over expelled and not justified?

   27. MSMati Staniszewski51:07

       Well, it's a risky business b- wh- where- where, you know, like there will be a winner from some of whether it's, uh, I love what LovaBall is doing with Anton, but Replit, vZero, Bolt are all incredible companies, too, um, or products too that are doing some- some- some- some great work. I do think at least one, maybe more than one will create something special. The market is so big that all of them can create something special. Um, but yes, there will be definitely a loser in that mix too, and- and- and that company and the margins that they are carrying does just not make long term, long term sense, and probably the same for coding ops. But I think it's a cool bet. I think it's c- co- c- n- not even like cool from the sense of like it's nice, uh- uh- uh, application, but it's like it's an ambitious bet that they are trying to win with the biggest companies in the world, that- that are trying, will try, or are trying with Google, with Firebase, um, uh, so f- uh, f- I don't remember the name of their product, but they are trying and the LovaBulls of the world are winning.

   28. HSHarry Stebbings52:02

       It goes to your point on brand. Neither of us can remember Google's product name. So i- I mean, it- it's just interesting.

   29. MSMati Staniszewski52:07

       Yeah. Anton is phenomenal market- mar- marketing person too. Like the founder led way he's doing is- is- is amazing.

   30. HSHarry Stebbings52:15

       One thing that's also really interesting is you were very horizontal in your customer from day one, which respectfully I would advise you very much not to be. (laughs) I think Philippe was telling me he was advising you the same. Be much more targeted in your ICP and have a clear mind of who that customer and user is, and you were much more broad and horizontal. And if you're advising founders, how do you tell them when they're launching from day one whether to be horizontal versus whether to be vertically specific?
7. 1:02:20 – 1:12:00

   Revenue Growth & Business Model

   1. MSMati Staniszewski1:02:20

      that's, that, that metric of course will, will change.

   2. HSHarry Stebbings1:02:22

      Can I ask what revenue is you at now?

   3. MSMati Staniszewski1:02:25

      When will this, uh, air? Or t- like, plus minus?

   4. HSHarry Stebbings1:02:27

      Uh, two weeks?

   5. MSMati Staniszewski1:02:29

      Two weeks. Two weeks. Okay. So, we cross 200 million now.

   6. HSHarry Stebbings1:02:33

      Whoa.

   7. MSMati Staniszewski1:02:34

      So it's a good number.

   8. HSHarry Stebbings1:02:35

      That's amazing. Yeah, that is a great w- per 250, if you've got that now.

   9. MSMati Staniszewski1:02:40

      Yeah, it's pretty good.

   10. HSHarry Stebbings1:02:41

       Fucking A.

   11. MSMati Staniszewski1:02:42

       Nice.

   12. HSHarry Stebbings1:02:43

       What was it-

   13. MSMati Staniszewski1:02:43

       The team is crashing.

   14. HSHarry Stebbings1:02:43

       What was it, end of '23?

   15. MSMati Staniszewski1:02:47

       The ... End of 2023, I think we were between 35... 35.

   16. HSHarry Stebbings1:02:52

       35 million?

   17. MSMati Staniszewski1:02:53

       35 million.

   18. HSHarry Stebbings1:02:54

       You went from beta launch in January.

   19. MSMati Staniszewski1:02:57

       To 35? We did, so we did 20 months to 100. And then, um ...

   20. HSHarry Stebbings1:03:03

       20 months to 100 million?

   21. MSMati Staniszewski1:03:05

       Yes. I think certainly that.

   22. HSHarry Stebbings1:03:06

       Fuck.

   23. MSMati Staniszewski1:03:07

       And then, um, around 10 months to 200. A bit longer.

   24. HSHarry Stebbings1:03:13

       (laughs)

   25. MSMati Staniszewski1:03:13

       15 months maybe.

   26. HSHarry Stebbings1:03:14

       Just crying (laughs) hearing these numbers.

   27. MSMati Staniszewski1:03:17

       (laughs)

   28. HSHarry Stebbings1:03:19

       Oh, it's so depressing. (laughs)

   29. MSMati Staniszewski1:03:21

       I mean, you know, the-

   30. HSHarry Stebbings1:03:22

       I'm so happy for you and I'm sad for you.
8. 1:12:00 – 1:14:43

   M&A Offers & Secondary Liquidity

   1. MSMati Staniszewski1:12:00

      would be like, "Oh, can we do... can we consider... are we open to M&A activity?" But it would be, um, a, a while back, and, um, um, yeah, I prefer also not to share.

   2. HSHarry Stebbings1:12:12

      Was it tempting?

   3. MSMati Staniszewski1:12:14

      It, it was tiny bit tempting, but in a, in a, you know, in um, in a way where we, um... In the f- in the first one or two examples. So the first one was when, when we were doing Series A. It was going really well. And then, of course, we like... you know, were unclear how, um, how, how this will continue. It was like the very beginning of the curve. And that, that... the first one was like, "Okay, this is first time we ever are doing this. Let's understand what, what they are, what they're offering." Um, but we were more interested into, into like seeing whether... m- more about the process itself, like how it happens, rather than actually, uh, giving, giving the company, um, because all, all... n- n- n- then it was clear, "Okay, they actually want to acquire us. This is what we understand, and we need to be part of the company." So we were, we were, we were a flat no, and since then, through any of the conversation now that we know that this is even approaching this M&A, oh, you know. I think it's, it's, it's... even more so now, I think, we are more-

Episode duration: 1:22:03