Inference by Sequoia Capital

Inference by Sequoia Capital

Inference by Sequoia Capital
Training Data
ElevenLabs’ Mati Staniszewski: Why Voice Will Be the Fundamental Interface for Tech
0:00
-59:53

ElevenLabs’ Mati Staniszewski: Why Voice Will Be the Fundamental Interface for Tech

Mati Staniszewski explains how ElevenLabs is creating the infrastructure for voice agents and real-time translation that could eliminate language barriers worldwide.
Jul 01, 2025
Share
Transcript
Post methodology: Claude 4.0 via custom Dust assistant @TDep-SubstackPost with the system prompt: Please read the text of the podcast transcript in the prompt and write a short post that summarizes the main points and incorporates any recent news articles, substack posts or X posts that provide helpful context for the interview. Please make the post as concise as possible and avoid academic language or footnotes. please put any linked articles or tweets inline in the text. Please refer to Podcast guests by their first names after the initial mention. Light editing and reformatting for the Substack editor.

ElevenLabs is changing the game for voice tech. In this episode, co-founder Mati Staniszewski explained how a childhood annoyance with monotonous dubbed films in Poland sparked a quest for natural, expressive text-to-speech. By staying laser-focused on audio—capturing emotion, context, and nuance—the team outpaced many broader AI models that were trying to do everything at once.

Their breakthrough came from understanding that audio presents unique challenges. Unlike text models that predict the next token, voice AI must capture subtle elements like sarcasm, timing, and emotional context. When someone says "what a wonderful day" versus saying "what a wonderful day," sarcastically, the entire delivery changes. This contextual understanding became their secret weapon.

The company's viral moments tell the story of their progress: from beta users copying entire books into their tiny text box to create audiobooks, to releasing the first AI that could laugh, to recent hits like Harry Potter by Balenciaga and Darth Vader's voice in Fortnite.

Mati envisions voice becoming the default interface for technology—transforming education with personalized tutors, enabling seamless cross-language communication, and powering everything from healthcare calls to customer support. The company is racing toward what he calls passing the “Turing test” for voice, where conversations with AI agents become indistinguishable from human interaction.

They're also tackling the dark side: impersonation risks. Every piece of content generated through ElevenLabs can be traced back to its creator, and they're developing detection models to identify AI-generated voice content across platforms.

Building from London with a fully remote team, ElevenLabs assembled top audio researchers globally—a smart move given the limited talent pool in voice AI. Their approach of keeping researchers close to deployment means innovations quickly reach users, creating a feedback loop that accelerates development.

The next frontier? Moving from their current “cascade” model (separate speech-to-text, language processing, and text-to-speech components) to a truly integrated duplex system that can handle real-time conversation with all the interruptions, emotions and nuances of human dialogue. Mati believes this could happen as early as this year.

As voice agents proliferate across industries, ElevenLabs is positioning itself not just as a voice provider, but as the infrastructure enabling a future where talking to technology feels as natural as talking to a friend.

Hosted by Pat Grady

Mentioned in this episode:

Discussion about this episode

User's avatar
© 2025 Inference by Sequoia
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture