What people would experience
ElevenLabs has introduced a service intended to let people speak different languages during the same live conversation. The company describes uses in telephone networks, customer-support calls and events, with translated speech delivered while the conversation continues. Its October 21 announcement limits initial access to enterprise customers through an ElevenAPI alpha.
The product information lists output in more than 90 languages, automatic language detection and streamed text that can support captions. Teams can adjust when translation begins, trading a longer wait for more sentence context.
An important voice distinction
The FAQ says translated audio does not exactly preserve the original speaker’s voice. A customer chooses an output voice or supplies a permitted voice clone. That is different from assuming the service automatically reproduces every caller’s identity. Complete pricing, regional eligibility and production terms were not established in the reviewed pages.
CGH perspective
For an initial evaluation, use conversations containing names, numbers, interruptions and corrections, then ask fluent reviewers to assess meaning rather than how pleasant the audio sounds. Participants should know when translation is active and have a way to repeat or clarify a statement. A call workflow also needs a fallback when translation fails or a language is unsupported. These are evaluation considerations; CGH has not tested the alpha or verified translation accuracy.