Mati Staniszewski

ElevenLabs reported $350 million of ARR at the end of 2025 and then added $100 million of net-new ARR in one quarter, putting John Collison’s rough arithmetic in the $450 million range. More than 50% of the business is now sales-led enterprise, with attractive entry pricing driving a “land and expand” motion from one workload into multiple departments.

The technical moat is a combination of model architecture and unusually rich data about how speech is delivered, not merely what was said. ElevenLabs built specialist labelling operations because public audio rarely identifies speakers, accents, emotions or prosody correctly; its internal transcription work subsequently became a customer product. The result is a model where even “Britishness is an emergent property,” rather than a hard-coded switch.

Voice AI’s near-term bottleneck is deployment and orchestration, not raw vocal quality. Staniszewski argued that production-quality asynchronous narration emerged three years ago, real-time speech two years ago and connected, context-aware applications only recently. Customer support can already pass a scoped “voice Turing test,” but open-ended gaming and general assistants remain far harder; ElevenLabs hopes the broader conversational-agent problem improves “in the next year or so.”

ElevenLabs is betting on cascaded agents—speech-to-text, LLM, then text-to-speech—because enterprises value reliability, visibility and control more than minimum latency. Direct speech-to-speech models are quicker but Staniszewski called them “definitely dumber”; they also lose visibility into intermediate pipeline steps and can be less reliable. Staniszewski expects a hybrid future: direct models for low-complexity interaction and cascades for higher-complexity actions, authentication and database retrieval.

The product opportunity extends well beyond synthetic voices into the full interaction layer between businesses and customers. ElevenLabs wants to supply models, knowledge connections, telephony, integrations, monitoring and safeguards across support, sales and marketing, while leaving deeply domain-specific applications to vertical vendors. Fixing voice also supplies the text layer, but complex multitouch reasoning and financial analysis are explicitly outside the company’s optimization target.

Voice models have structurally lighter training economics than frontier LLMs, but distribution and inference still matter. Leading voice models are “a few billion to tens of billion” parameters versus hundreds of billions for leading LLMs; customers pay mainly per token or minute, with annual commitments and volume discounts. ElevenLabs often offers a new, costlier model near cost so customers reveal its use cases and failure modes—turning adoption into both distribution and research feedback.

The next product frontier is individualized understanding and controllable delivery. ElevenLabs expects person-specific transcription in a forthcoming version, “hopefully in the next month,” supplementing diarization, noise reduction and keyword dictionaries; V3 already adds instructions such as slowing down or inserting dramatic pauses. Its expressive mode can also detect stress and generate a reassuring response in both language and tone.

ElevenLabs’ operating model is as important to the growth thesis as its models: 470 employees, product and research initiatives generally below 10 people and more than 15 direct reports for each founder. Nontechnical functions receive technical leads who automate workflows, while self-serve access exposes the newest technology to developers before it is enterprise-ready. Staniszewski’s cultural filter is “agency”; Collison’s sharper formulation was that high-agency people win from AI while low-agency people inside organizations lose out.

ElevenLabs reached $300 million ARR with 350 employees by pairing a five-million-MAU creator product with an enterprise agents business approaching half of revenue. Self-serve subscriptions and creators remain roughly 50% of the mix; several thousand enterprise customers range “from Fortune 500s to some of the fastest-growing AI startups.” The operating signal is breadth without abandoning the original research advantage.

The founding wedge was Poland’s broken dubbing experience, where one flat narrator voices every character, but the thesis expanded into voice as computing’s natural interface. ElevenLabs aimed to carry the “original voice, original emotions, original intonation” across languages, as demonstrated with Lex’s interview with Narendra Modi. Mati’s larger claim is that keyboard-and-screen interaction “feels broken,” making the opportunity far larger than conventional dubbing spend.

Mati’s operating model is research first, product in parallel, with small cross-functional labs formed around concrete problems. A roughly five-person voice lab solved human-sounding narration before ElevenLabs expanded into audiobooks and dubbing; an agent lab then combined speech-to-text, LLMs, text-to-speech, integrations, testing, and monitoring. Product teams see the research roadmap and generally build alternatives only when the relevant breakthrough looks more than three months away.

Customer support is the first major agent category, but voice is moving from reactive ticket resolution into discovery, checkout, education, media, and government services. Meesho’s agent progressed from refunds and tracking to product guidance and potentially checkout; MasterClass lets learners practice a negotiation with Chris Voss; Ukraine is pursuing what Mati calls a “first agentic government.” That expands the economic case from labor savings toward revenue generation and entirely new experiences.

Voice quality is not one scalar benchmark because preference can flip with voice identity, language, audience, and delivery even when underlying model quality differs. One customer requested a voice “as robotic as possible,” while a Japan-and-Korea deployment wanted an excitable voice for younger callers and a calm, slow one for older users. ElevenLabs therefore supplies voice coaching and expects selection eventually to become dynamic for each person and context.

ElevenLabs assumes base models will commoditize and treats research leadership as only a six-to-12-month advantage. Mati says “research is a head start”; durable value must accumulate through distribution, voices, integrations, workflows, and a product layer that connects models to business logic. Open-source narration is already strong, but controllability and real-time orchestration remain differentiated, while a concentrated pool of perhaps 50–100 elite audio researchers—Mati says roughly 10 are probably at ElevenLabs—can still produce architectural leaps.

The near-term roadmap targets controllable multimodal creation and lower-latency, emotionally aware agents while retaining cascaded systems for reliable enterprise work. Scribe v2 is reported at under 150 milliseconds and 93.5% accuracy across the top 30 languages on FLEURS; fused speech-to-speech may become more expressive but can introduce hallucinations and offers less visibility. Sarah estimates that humanlike conversational interaction is still at least a year away but may arrive within a year, with real-time dubbing or translation within two; personalized education—“your own teacher on demand”—is the largest unrealized application.