Imagine calling a contact center in your own language and hearing the agent’s reply — in their real voice — in yours. That’s the pitch behind ElevenLabs Realtime Translation, a new streaming voice layer that turns live speech into translated audio as people talk.
What it does
The product targets telecom operators, contact centers and live events. Each person speaks their own language and hears the other in theirs, across 90+ languages. Instead of caption-only subtitles or a stitched pipeline of separate speech-to-text, LLM and text-to-speech steps, it goes from live speech to expressive translated audio over a single WebSocket connection.
What makes it different
The headline feature is voice consistency: the translated audio keeps one consistent voice — either a voice from ElevenLabs’ library or your own cloned voice. That means a German-speaking agent talking to an English-speaking caller sounds like themselves the whole time, not like a robot narrator.
It also streams captions and transcripts alongside the audio, and lets operators tune the speed-vs-accuracy tradeoff per use case — a fast-and-loose setting for live events, a careful one for legal or medical-adjacent calls.
Who it’s for
This isn’t a consumer app — it’s infrastructure for businesses that run multilingual conversations at scale. The product is listed as launching in 2026 with free options available; pricing beyond that wasn’t stated on the launch page, so check elevenlabs.io/realtime-translation for current terms before evaluating.
Also Read: find more AI tool launches in our AI News section
Source: Product Hunt — read the full story there.
