ElevenLabs CEO Mati Staniszewski says the company has passed the uncanny valley in call centres and narration, but not yet in conversational settings like gaming or personal AI.
ElevenLabs runs two lines, an agents platform for conversational AI and a creative platform for narration, voiceovers, dubbing and music, with revenue close to fifty-fifty between consumer creative use and enterprise agents.
Staniszewski predicts the dramatic shift to voice AI will take five to ten years because adoption and business process change are slow, while Kleiner Perkins partner Joubin Mirzadegan remains skeptical about enterprise uses beyond coding and support.
ElevenLabs bets it can out-build OpenAI to own voice
ElevenLabs is betting that voice becomes the main interface to technology and that it can be the company providing it. Synthetic speech already matches or beats IVR menus and stretched human agents in support and narration. That lets one company sell creative tools to consumers and conversational agents to enterprises, provided it keeps out-building OpenAI and Gemini on models while shipping a full application platform.
Takeaways
Revenue splits evenly between creators and enterprises. Staniszewski says revenue is close to fifty-fifty. Consumers use the creative platform for narration, voiceovers, dubbing and music. Businesses use the agents platform for conversational AI, which ElevenLabs also runs on its own website, product and customer support.
Voice agents produce measurable business gains. An Italian real estate company raised its lead qualification rate from 20 to 60 percent after deploying an agent. A Japanese healthcare support company tuned voice and response style differently for older and younger callers.
ElevenLabs must out-build frontier labs on two fronts. Staniszewski says the company needs models better than OpenAI and Gemini, and claims it already leads in text to speech, speech to text and music. It also needs an application platform integrating with Salesforce and ServiceNow. Mirzadegan compares this to Cursor surviving OpenAI and Anthropic.
The idea came from Polish single-voice dubbing. Foreign films in Poland are often narrated by one voice for every character. The founders, high school friends, started in 2022 with remote researchers hired wherever the best were. Early breakthroughs were emotional speech and voice recreation, when most synthetic speech still sounded robotic.
Small autonomous teams drive rapid scaling. ElevenLabs grew from under 100 people to 330 in about a year, with hubs in London, New York, San Francisco and Warsaw, and expects over 500 within twelve months. About 20 teams of five to 10 people make most decisions themselves.
A secondary tender keeps staff aligned long term. Mirzadegan cited a $6.6bn figure and a $100m secondary-only tender. Staniszewski says it gives early liquidity while preserving incentives for bigger outcomes. He talks of a $121bn ambition and a generational company that outlives its founders.
Where host and guest disagree
Staniszewski says ElevenLabs has passed the uncanny valley in call centres and narration, though not yet in gaming or personal AI, and is 99.9% confident it will be crossed in most cases.
Mirzadegan says he can almost always tell AI audio and text, prefers knowing he is talking to a machine, and cares mainly about fast resolution. He argues ElevenLabs can become huge without crossing the valley.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Mirzadegan imagines a director handing an LLM a script for a Goodfellas-style film. Staniszewski expects AI to replace parts of moviemaking middle to middle and says writing such a script would be extremely hard.
Mirzadegan doubts enterprise AI beyond coding and support has worked well. Staniszewski is maximally bullish but expects the big shift to take five to ten years and does not see AGI in the next two or three.