- ElevenLabs' unique approach to data and model architecture creates human-like, emotionally nuanced AI voices.
- Despite rapid technological advancements, a 'deployment gap' hinders widespread adoption of cutting-edge voice AI in consumer products.
- The company achieved a staggering $100 million in net new ARR growth in a single quarter, driven by enterprise adoption and a self-serve strategy.
Mati Staniszewski, co-founder of ElevenLabs, pulls back the curtain on the groundbreaking innovations propelling his company to the forefront of the voice AI revolution, discussing everything from the nuances of audio model architecture to the strategic decisions behind their meteoric rise.
ElevenLabs has rapidly scaled to become a leader in AI audio by focusing on capturing the 'humanness' of speech. Staniszewski explains that their core innovation lies in applying transformer and diffusion models to the speech space, enabling the prediction of the next phoneme (the smallest unit of sound) and incorporating context to generate highly realistic and emotionally inflected voices. Unlike traditional methods that rely on hard-coded parameters, ElevenLabs' models deduce characteristics like accent and emotion as 'emergent properties,' leading to unparalleled naturalness. A significant part of their success stems from a massive investment in proprietary data labeling, even developing their own speech-to-text models when market solutions fell short.
Despite the advanced capabilities of modern voice AI, Staniszewski highlights a significant 'deployment gap.' While the technology is ready, industries like automotive and consumer applications are slow to integrate it, leaving users with outdated voice experiences. He notes that true conversational AI, capable of passing a 'voice Turing test' with human-like interaction and complex orchestration, remains a hard research problem, though ElevenLabs is making strides in specific domains like customer support. Looking ahead, the company is poised to launch person-specific transcription, allowing AI to learn and perfectly understand individual voices, a critical advancement for fields like healthcare and personalized device control.
ElevenLabs' business model combines foundational audio model development with a platform for businesses, focusing on horizontal applications rather than niche verticals. This strategy, coupled with a strong self-serve component, has fueled explosive growth, with the company announcing an astonishing $100 million in net new Annual Recurring Revenue (ARR) in a single quarter. Staniszewski attributes this success to a belief in their technology's value, offering attractive economics for adoption, and a 'land and expand' approach within enterprises. Internally, ElevenLabs operates with small, flat teams and a culture that emphasizes agency, technical proficiency across all departments, and a commitment to solving complex, real-world problems, as exemplified by their work with Ukraine's tech-forward government.
The impact of ubiquitous voice AI, Staniszewski predicts, will be profound, breaking down language barriers (like improved dubbing and real-time translation), restoring voices to those who have lost them, and enabling a new generation of proactive and reactive AI agents. From calling pubs to check Guinness prices (the 'Gindex') to powering lead generation for banks, voice agents are transforming how businesses interact with the world. ElevenLabs' commitment to continuous innovation, a user-first self-serve strategy, and an agile, AI-native organizational structure positions them to lead this transformative shift, proving that the future of communication will indeed be heard.
“The technology in many of those cases is ready. There's a deployment gap to what you are saying is like an automative or some of the big companies are not adopting that quickly enough or bringing that into the production.”
- Mati Staniszewski, Co-founder of ElevenLabs
