ElevenLabs CEO Mati Staniszewski on Voice AI, Growth and Building from Europe

Harry Stebbings interviews ElevenLabs co-founder and CEO Mati Staniszewski about building an AI voice company, fundraising, hiring and growth. Staniszewski speaks for the business he leads, so product comparisons, financial figures and forecasts are company claims unless independently supported. The host says he and the company have worked on projects together.

Staniszewski describes growing up near Warsaw and meeting his co-founder in a competitive high-school environment. He says being surrounded by motivated people broadened his sense of what was possible, and that the founders now seek a similarly talented team.

The founders knew each other before starting ElevenLabs and experimented together while working at Google and Palantir. A project analyzing speech led them toward audio. They were also struck by Polish movie dubbing in which one flat voice narrates many characters; that experience helped inspire their initial idea.

They divided early work between research and customer discovery. His co-founder explored whether existing technology could create better dubbing, while Staniszewski emailed creators. He recalls a roughly 15% response rate to early personalized outreach but says interest in dubbing was not yet urgent.

Conversations suggested creators wanted simpler tools first: correcting spoken mistakes, hearing scripts and generating voiceovers. Staniszewski says the team shifted toward text-to-speech and built a model aimed at more expressive narration. The description of model quality is the company’s account.

Asked whether an independent model was necessary, he says available speech models in 2022 did not meet their needs. Today he sees a choice between separate modality models and approaches combining reasoning and speech. Whether to build or use another provider depends on the current task and available technology.

Stebbings asks whether model progress is plateauing. Staniszewski says improvements for some narration tasks may become less dramatic, while expecting further progress in voice; he also accepts that large-language-model progress may be flattening to some extent. These are assessments and predictions, not measured scaling laws.

Stebbings asks why OpenAI cannot build the same product. Staniszewski expects it to compete and argues ElevenLabs can differentiate through focused research, execution and product workflows. He cites performance claims and the strength of his team, but the interview supplies no common independent benchmark for every use case.

On the competition for researchers, he emphasizes the opportunity to deploy research quickly, work with a small team and share in ElevenLabs’ upside. His account of the company’s attractions and other firms’ processes is a founder’s perspective, not evidence about every employee’s experience.

The pre-seed raise was difficult, he recalls: investors questioned the research, the size of the market and defensibility. He says the founders spoke with roughly 30 to 50 investors and initially used their savings to fund GPUs and the first hires.

He says the company raised $2 million in the pre-seed process and recalls a $9 million post-money valuation for the first investment, with others layered in. He distinguishes completing the round around the third quarter of 2022 from announcing it with the beta in January 2023. These are his recollections, not independently checked transaction documents.

The initial dubbing outreach did not show strong product-market fit, he says. After a shift toward narration, a blog post brought a waiting list and early testers found practical uses. One audiobook author reportedly assembled a book from many short generated clips and shared it with others; this is an anecdote, not evidence of platform-wide retention.

Staniszewski is cautious about declaring product-market fit from an enthusiastic launch. He says lasting value over five to ten years matters more than a single early signal, while acknowledging that the team saw momentum after release. Long-term retention and outcomes were not established by those anecdotes.

He prefers tying fundraising announcements to product or customer news and focusing launches on reaching actual users. He recalls a traditional press announcement having limited immediate impact while newsletters, Discord, Reddit and Hacker News brought more attention. The relative effect is his experience, not a universal marketing rule.

He advises founders to avoid continuous fundraising when it distracts from users, but to maintain relationships with a few potential investors. Specific requests for introductions or hiring help can test whether an investor is useful before accepting money. Stebbings agrees that early help can reveal genuine interest.

For angels, he values missing domain expertise, access to relevant communities and go-to-market experience. These criteria reflect ElevenLabs’ circumstances rather than a prescribed investor selection formula.

The speakers discuss the 2023 round involving Andreessen Horowitz and NFDG. Staniszewski recalls Nat Friedman testing the API and providing specific feedback, and says Andreessen Horowitz made introductions before investing. His praise concerns existing backers with financial interests in the company.

He values speed in investment decisions and product execution. Stebbings notes that some founders run longer fundraising processes, creating tension for an investor who must decide quickly. Both discuss the trade-off from their respective founder and investor perspectives.

Staniszewski describes organizing work into relatively small teams with substantial autonomy. Teams can focus on product areas such as core experience, studio or voice agents, with distinct functions supporting them. This is his account of the company’s organization at the time.

He recalls tension between research, engineering and go-to-market teams when an enterprise customer assembled ElevenLabs components into a dubbing product and launched before ElevenLabs. He says the team had delayed its own launch while seeking a more polished experience.

He does not name the customer and says he does not think the customer timed its launch based on ElevenLabs’ plans. He recalls the customer describing it as a quick project and claims it generated tens of millions of dollars in revenue. That figure and the sequence are his account, not independently verified results.

His lesson for a founder is first to acknowledge the frustration and examine what happened, then move quickly to a plan rather than pretend nothing went wrong. He says the team should learn from a mistake without repeating it. This is his management interpretation of one incident.

On competition, Staniszewski says research, product and distribution all matter. He estimates ElevenLabs has a six-to-twelve-month research lead depending on the use case, but that is a company executive’s judgment, not an independently established benchmark.

He says ElevenLabs built data centers for training while using partners for inference. The company expected ownership to break even against renting over roughly two years, depending on training needs and hardware improvements. He says the decision worked for them but acknowledges future technology could change the equation; no financial model is supplied.

Stebbings raises poor unit economics at AI application companies. Staniszewski says some firms are betting on future cost improvements and customer loyalty, while acknowledging that some may fail. He claims ElevenLabs’ unit economics are healthier than those of many peers without disclosing comparable figures.

On horizontal versus vertical products, he says a broad approach can make sense when a new technology’s full customer base is not yet clear, whereas existing domain expertise may support a focused vertical. He presents this as a contextual decision, not a universal rule.

He explains why ElevenLabs moved away from traditional job titles: small teams can change quickly, and he wants decisions and leadership to depend on impact rather than tenure. He says teams still have decision-making leads, whose roles are not necessarily permanent.

Asked whether European companies lack experienced executives, he says ElevenLabs aims to grow talent locally and pair employees with advisers from its US investor network. He agrees building in Europe can be harder in some respects but argues the region has strong talent and companies should seek global markets.

He says the company now receives more inbound interest from candidates and that joining a well-known company is not inherently a bad motivation. On hiring mistakes, he wishes he had acted sooner in some cases where a hire was not working. These are management views, not a complete account of employees’ experiences.

Staniszewski expects roughly 400 employees by year-end, up from about 250 at the time, and describes plans to expand local teams in markets including Brazil, Japan, India and Mexico. The headcount and hiring pace are company-reported; the year-end figure is a plan, not a verified outcome.

He says revenue per employee matters over the long term, but hiring can temporarily reduce that metric while supporting distribution and retention. His argument depends on future productivity and customer outcomes, not a guarantee that every hire will pay for itself.

Staniszewski says the company has crossed $200 million; the episode description identifies the metric as annual recurring revenue, or ARR. ARR is an annualized recurring-revenue measure, not realized annual revenue, net profit or cash flow. The interview does not provide audited financial statements.

The speakers recall approximately 20 months to reach $100 million and then discuss the time to $200 million. Staniszewski mentions around ten months and also suggests it may have been closer to fifteen, so the exact interval is unclear in the conversation. He adds that revenue can decline as quickly as it grows.

He says conversational agents are now a major focus and reports the largest contract is around $2 million, mainly describing call-center, support and assistant applications. He does not specify the contract’s period or recognize it as annual recurring revenue in that exchange. Named deployments and self-serve activity are company statements.

Stebbings asks if ElevenLabs can reach $300 million in five months. Staniszewski says he would like to break the record if the revenue is healthy and the work valuable. That is an aspiration, not a forecast or an achieved milestone.

The speakers discuss a $3.3 billion financing valuation announced in January 2025. Staniszewski says the round was done toward the end of 2024 and gives approximate, varying revenue figures for the time the documents were prepared and signed. The valuation is not revenue or a realized return, and the interview does not reconcile those estimates.

He says new funding can bring forward investment in models, international expansion and enterprise agent capabilities. The test, in his view, is whether a new effort can run in parallel without distracting the core business. More capital does not by itself prove those investments will generate returns.

He believes voice agents could eventually generate multibillion-dollar revenue if the company executes well, potentially expanding into other customer-support channels. ElevenLabs has a public partnership with Decagon, he says, while acknowledging the companies could overlap more as strategies evolve. His revenue scenario is speculative and commercially interested.

Asked about employee resistance to agents, he describes a division in which AI handles routine tasks and people focus on specialized work. He uses scheduling and refunds as examples conditioned on safeguards and authentication, and says sensitive patient navigation still requires human expertise. These are illustrative views, not evidence that any deployment is safe or that employment effects will be benign.

Staniszewski says recognizable investors including Andreessen Horowitz and Sequoia helped the company gain credibility with customers. His favorable assessment concerns the company’s own backers. He also values technical investors’ advice and reputation among engineers.

He says ElevenLabs has received acquisition approaches, generally informing investors and examining the basics. He recalls early conversations being somewhat tempting but says the founders preferred to continue building independently. He does not disclose a verified purchase price or completed transaction.

He says founders and employees with vested stock can access secondary sales or tender offers in many rounds. In his view partial liquidity can let people take long-term risks while meeting personal financial needs. The interview does not supply quantities, terms or independent transaction records.

In quick-fire questions, he predicts voice will become a primary interface for much technology and says he believes globally scaled companies can be built from Europe. Both are his views of future adoption and company building.

He says the company has become more willing to explore products using outside research rather than requiring all underlying research to be internal. Asked about rapid ARR growth as a metric, he says its importance depends on the time horizon. Neither comment is a forecast of ElevenLabs’ future growth.

On European AI policy he proposes following US approaches or creating an opt-in jurisdiction, while acknowledging significant repercussions. This is a policy preference, not a legal analysis. He also worries that an excessive founder brand could obscure the contributions of researchers, engineers and other teams.

He closes by recalling advice about the risk of not taking risks. He mentions considering a large acquisition costing hundreds of millions of dollars, while saying the company believes it may be able to build internally. No target or transaction is confirmed in the interview.