You're right to check — use the turn detector with Speechmatics and keep VAD. Quick guidance and a minimal snippet below. • Use LiveKit's turn detector (it combines STT transcripts + VAD silence events) for better end-of-utterance decisions. • Configure Speechmatics with

end_of_utterance_mode="NONE"

so Speechmatics does not do endpointing that conflicts with LiveKit's turn detector. • Keep VAD (e.g., Silero) for low-latency silence/interruption detection — don't remove it unless you have another reliable low-latency source of silence events. Sample Python snippet:

from livekit.agents import AgentSession
from livekit.plugins import silero, speechmatics
from livekit.plugins.turn_detector.multilingual import MultilingualModel

session = AgentSession(
    turn_detection=MultilingualModel(),
    vad=silero.VAD.load(),
    stt=speechmatics.STT(
        end_of_utterance_mode="NONE",
        # language="en",
        # enable_partials=True,
    ),
    # tts=..., llm=...
)

Would you like a version of this snippet for Node/JS or help tuning VAD sensitivity? Sources: LiveKit turn detector plugin | LiveKit docs | Turn detection and interruptions | LiveKit Docs | Speechmatics STT plugin guide | LiveKit Docs