Speak in your own language, hear the other party in your own language
Voice AI company ElevenLabs said on its official X account on Oct. 21 (early Oct. 22 Japan time) that it had launched "Eleven Realtime Translation," which interprets conversations during calls on the spot. Each party speaks in their own language and hears the other party in a natural voice in their own language. It is first being offered as a limited trial version to business customers that use the company's API.
It supports more than 90 languages and automatically detects the language being spoken. It is built on "Eleven v4 Turbo," a speech synthesis model with reduced latency. How quickly translation begins can be adjusted, letting users choose between speed and translation quality.
The interpreter's voice stays the same throughout the conversation
The interpreter's voice can be chosen from more than 17,500 voices or created by imitating the user's voice. The chosen voice is kept the same throughout the conversation, speaking in the same voice even when the language changes.
As uses, the company cited examples such as telecom carriers building it into calling features, placing it between agents and customers at call centers, and letting attendees at events listen in their own language. Germany's Deutsche Telekom already uses it in a feature of its call assistant, "Magenta AI Call Assistant." Pricing was not disclosed.
ElevenLabs had released "Eleven Realtime v1," a new model for agents that converse by voice, just the day before, on Oct. 20.
