ElevenLabs launches realtime voice translation in alpha
ElevenLabs has released Eleven Realtime Translation, a live speech-to-speech model for enterprise API customers, aimed at telecom operators and contact centers that want to serve callers in many languages without human interpreters.
ElevenLabs announced Eleven Realtime Translation, a live voice translation model now in alpha for enterprise customers of its ElevenAPI platform.
The system lets each participant speak their own language and hear the other person in theirs while the conversation is happening. ElevenLabs is pitching it for three settings. Telecom operators can build it into the calling layer of their networks. Contact centers can place it between agent and caller, so one agent can handle more languages without bringing an interpreter onto the call. At live events, attendees can hear a speaker in their own language.
Deutsche Telekom is an early user. It is building new calling features for its Magenta AI Call Assistant with the model. "It's really a complete new era of how voice can actually work and connect people, and also break down barriers," said Lena Drubel, VP Consumer AI at Deutsche Telekom.
By default the product runs on Eleven v4 Turbo, the company's low-latency speech model for realtime use. It covers more than 90 languages and detects the spoken language automatically. Customers can pick any of more than 17,500 voices in the ElevenLabs library or use an Instant or Professional Voice Clone. The chosen voice stays consistent across languages for the whole session.
Deployments can trade speed against quality. Teams decide how quickly translation starts, choosing more accurate output for complex calls such as billing or claims, or the lowest latency for faster back-and-forth.
ElevenLabs did not publish pricing, latency figures, accuracy benchmarks or a timeline for general availability.