Speech-to-text transcript

Stable per-event idempotency key; identical across delivery retries and across all subscribers of the event.

Instance identifier

Leg identifier

Room identifier

Which of the leg's audio streams this came from. Empty when the leg's audio is the call itself; set for a recording session, where each stream is a different party. Resolve it through GET /v1/legs/{id}/siprec

Transcribed text

Whether this is a final or partial transcript

Whether the speaker stopped talking, as opposed to is_final's 'this segment will not change again'. Deepgram only; always false for providers that do not report it

Where in the stream this was said, in milliseconds from the first audio the transcriber was given. Absent when the provider reports no timing. Not the same as the event's arrival time, which is when the provider finished rather than when the words were spoken

End of the span audio_start_ms opens