Introducing Eleven Realtime: a simpler way to build natural voice agents
ElevenLabs launches Eleven Realtime, an offering for building voice agents, with the Mini and Max v1 models available in ElevenAgents.
What the article says
ElevenLabs launches Eleven Realtime, two speech-to-speech models for voice agents. Building a natural voice agent normally means choosing and tuning speech recognition, voice activity detection, turn-taking, an LLM and text-to-speech separately; ElevenLabs says it has tuned the whole stack end to end and makes some judgment calls adjustable in ElevenAgents.
Both models support more than 90 languages and over 17,000 voices: v1 Mini, for natural low-cost conversations, and v1 Max, for complex tasks such as customer support and lead qualification. According to Artificial Analysis, v1 Max answers 98.8% of Big Bench Audio questions correctly, with 0.57 seconds to first audio. In ElevenAgents, users pick the model from a dropdown; through the Realtime API, which follows OpenAI's Realtime specs, developers switch by changing the endpoint and model name, but only over WebSocket and without platform features such as channels or knowledge bases.
Read the official article AnnuIA profile
Summary written by AnnuIA in its own words from the official page of ElevenLabs. The content and rights belong to the publisher: please read the original article.