ElevenLabs has launched an alpha version of Realtime Translation for enterprise customers. The service translates both sides of a spoken conversation as it happens, letting each participant speak and hear the call in a language they understand.
The product uses the company’s low-latency Eleven v4 Turbo speech model and supports automatic language detection across more than 90 languages. A deployment can use one of more than 17,500 library voices or a custom voice clone, keeping the selected voice consistent across languages.
Operators can trade response speed for translation quality. ElevenLabs suggests lower latency for quick back-and-forth exchanges and more processing time for complex calls involving billing or insurance claims. Proposed uses include contact centers, telecom calling features and live events.
Deutsche Telekom is already building the technology into features for its Magenta AI Call Assistant. Broader access has not been announced: the current release is an alpha available through ElevenAPI to enterprise customers, so performance and language quality should be evaluated in the specific accents and high-stakes conversations where it will be used.