Member Junction
    Preparing search index...

    One frame of raw per-participant audio the seam surfaces for diarization + the agent's "hearing". Because LiveKit subscribes tracks per participant, every inbound frame is already attributed to a single speaker — diarization comes free from the SFU.

    interface LiveKitAudioFrame {
        DisplayName?: string;
        ParticipantIdentity: string;
        Pcm: ArrayBuffer;
        TimestampMs?: number;
    }
    Index

    Properties

    DisplayName?: string

    The participant's display name at capture time, when known.

    ParticipantIdentity: string

    The LiveKit participant identity this audio came from (the diarization speaker label).

    Pcm: ArrayBuffer

    Raw PCM audio bytes for this frame.

    TimestampMs?: number

    Optional epoch-ms capture timestamp.