ElevenLabs CEO, Mati Staniszewski, aims to make voice AI conversational to pass the Turing test this year or in early 2026.
The company is transitioning from a cascading architecture to a unified duplex model for voice AI, balancing expressiveness and reliability.
Challenges such as latency and integrating audio with large language models persist, but ElevenLabs is optimistic about overcoming them.
The CEO believes voice will become a primary interface for technology interactions and recently launched Eleven v3, a text-to-speech model with enhanced features.