Nikhil Kamath says his side recently joined Nothing's cap table, and Nothing's sales rose from 500m last year to about 900m this year.
Mati Staniszewski argues voice will become a major interface once voice quality nears human level, memory and knowledge access improve, and a suitable form factor arrives, with behind-the-ear headphones his likely pick.
Mati Staniszewski says ElevenLabs is roughly half creator and half enterprise, and that it does not depend on OpenAI for its voice business even as OpenAI builds toward a personal assistant.
ElevenLabs bets interactive voice agents will dwarf the dubbing market
ElevenLabs is betting that voice becomes the main way people interact with technology. Mati Staniszewski argues the larger prize is interactive agents that answer calls, sell and teach, well beyond dubbing. That shift still needs near-human voices and better wearables, and he believes ElevenLabs can lead it without depending on OpenAI.
Takeaways
Interactive agents dwarf dubbing as a market. Nikhil cites research sizing dubbing under $1bn, rising to $3bn in a decade. Mati argues interactive uses are bigger. About 500 companies in India and over 10,000 worldwide build voice agents, and an AI SDR flow lifted TVS Motor's satisfaction and lead quality.
Voice interface awaits better models and hardware. Mati expects voice to become a major interface once quality nears human level and assistants gain knowledge and memory. Devices will blend wearables such as headphones and glasses, and he would pick behind-the-ear headphones. He sees Nothing as a possible first AI-native device.
A business split between creators and enterprises. ElevenLabs is roughly half creator, half enterprise. Its creative platform handles narration and localization for podcasters and media companies like Star. Its agents platform serves customer experience and education, with Meesho automating support. Masterclass uses it to recreate Gordon Ramsay and Chris Voss as AI teachers.
Dubbing is cheap but nuance needs people. Automatically dubbing an hour-long podcast costs a few dollars, yet the Lex Fridman and Narendra Modi conversation needed humans checking each translation. New models keep the speaker's voice and emotion, though word order matters, since German nouns shift where a smile lands. Lip animation is expected soon.
Pair domain expertise with traditional industries. Mati points entrepreneurs at sectors such as automotive and healthcare. A voice business for BMW would start with customer support. Meesho, which recently listed, handles about 60,000 calls a day, showing the volume older businesses can automate.
No AI bubble, except GPU resellers. Nikhil says ElevenLabs has raised roughly $300 to $400m and is worth a few billion dollars. Mati sees no overall bubble but doubts GPU providers built on NVIDIA. He hopes companies anywhere can earn Silicon Valley valuations if they build globally.
Network effects survive a more national internet. Nikhil predicts countries will build their own versions of global products, as Russia and China did. Mati agrees data should stay local and expects open-source base models adapted with national data, but says network effects hold. ElevenLabs' marketplace of 10,000 voices has paid creators over $11m.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Designing an algorithm-free social app
Nikhil wants an Indian social app not governed by an algorithm that triggers hate or greed, and blames Snapchat, Instagram, bots and trolls for today's feeds.
Entry would run through a tightly controlled funnel of verified real people.
Mati suggests voice-based feeds and assistants, including hearing essays in the author's own voice, and cites AirChat as a voice social precedent.
He advises avoiding transactional incentives and rewarding authenticity over outrage.