ElevenLabs Launches v4 Speech Models With 90-Language Support and Lower Latency
ElevenLabs’ new architecture can clone a voice from as little as 10 seconds of audio, preserve identity across longer passages, and adapt expression to context. Support expands from 70 to 90+ languages, while v4 Turbo targets voice agents.
ElevenLabs released two new speech models on Monday, September 28, 2026 — ElevenLabs v4 and v4 Turbo — adding support for more than 90 languages, expanded expression controls, and reduced latency aimed at voice agent applications.
The v4 generation introduces a new architecture that enables faster voice cloning, requiring as little as 10 seconds of audio to replicate a voice. The model also maintains voice identity more consistently over longer text passages and adjusts expression based on the context of what is being read. Building on the inline expression tags introduced in v3, v4 allows users to stack multiple tags and have the model follow them in sequence.
Language support has grown from 70 in the previous version to more than 90 in v4. ElevenLabs said the largest quality improvements were observed in Japanese, Brazilian Portuguese, Mandarin, and Cantonese.
The company highlighted v4’s suitability for voice agents, noting that the model can begin generating audio as soon as the underlying language model starts producing a response. It also handles confrontations, escalations, and holds differently to support better issue resolution — a relevant capability given that more than 55% of ElevenLabs’ business now comes from large enterprise customers.
ElevenLabs faces a growing field of competitors, including Cartesia, Deepgram, Fish Audio, Boson, and WellSaid Labs, as well as voice model improvements from Google and OpenAI.
The company raised $500 million from Sequoia earlier in 2026 at an $11 billion valuation, and reports suggest a follow-up round could value the company at $22 billion. Its annualized revenue run rate has risen from roughly $330 million at the start of the year to over $600 million. Headcount has surpassed 800, with hiring concentrated in markets including India, Europe, and Brazil.
Co-founder and CEO Mati Staniszewski told TechCrunch the company is targeting an IPO “in the next years,” without committing to a specific timeline.