Realtime Translation does not carry the original speaker's voice across languages exactly. Instead, you select the output voice from a library of 17,500+ voices, or use an Instant or Professional Voice Clone as the voice_id to approximate a specific speaker.

Yes. Both parties speak their own language and hear the other in theirs. Bidirectional is configured at the network or product level, for example by connecting each caller to a dedicated instance of the endpoint.

With the eagerness parameter (low, medium, or high), set per session. High starts translating soonest for the lowest latency. Low waits for more of the sentence for higher-quality translation, with more delay. Medium is the default. 

Eleven v4 Turbo is the default. Customers can also set tts_model_id to eleven_flash_v2_5 or eleven_v3. Eleven v4 Turbo gives one voice stronger performance across multiple languages.

Yes. Every session streams translated text alongside the audio, plus interim and final transcripts of the source audio. You can use these to power live captions or logging on top of the audio.