ElevenLabs launches Eleven Realtime speech-to-speech models for voice agents
By Steven Van ·
Eleven Realtime v1 Mini and v1 Max each work as a complete voice agent, starting to talk in about half a second in 90+ languages.
ElevenLabs introduced Eleven Realtime on October 20, a new class of speech-to-speech models for voice agents. ElevenLabs says voice agents usually make you choose between quick replies and precise answers, and Eleven Realtime is built to give you both.
Each model is a complete voice agent, with speech recognition, turn-taking, the language model and the voice tuned together as one system. There are two: v1 Mini, the lower-cost option for assistants, tutors and interactive characters, and v1 Max, which reasons before it responds and is aimed at customer support, lead qualification and multi-step tasks with tool calls.
ElevenLabs says v1 Max starts speaking in 0.57 seconds on average and v1 Mini in 0.62 seconds, and that v1 Max answered 98.8% of the spoken reasoning questions in the Big Bench Audio test correctly. Both support 90+ languages and voices from the ElevenLabs library.
Eleven Realtime is available in ElevenAgents and through a Realtime API that is compatible with OpenAI’s Realtime API, so developers can switch by changing the endpoint and model name. The announcement does not list per-minute pricing.
Sources: ElevenLabs’ announcement and demo, the launch blog post and the Eleven Realtime page.