Thanks — good question. I’m responding to your two points: (1) why the docs warn against using Krisp Front End together with Noise Cancellation (NC) or BVC, and (2) whether

krisp_enabled=True

on a trunk is the same as “Krisp Front End.” Short answer • The warning exists because using Krisp on the front end and also applying NC/BVC later causes double-processing of the same audio, which can degrade speech quality and hurt STT accuracy. • Yes — setting

krisp_enabled=True

on a SIP trunk is equivalent to enabling the “Krisp Front End” (server-side Krisp) for that trunk. If you enable that, you should not also run NC() or BVC() on the agent side. Why this matters (brief) • Noise suppression algorithms are designed to work on raw audio. If audio is pre-processed by Krisp on the trunk and then processed again by additional NC/BVC, the second pass sees already-filtered audio. That can remove speech content (especially beginnings of words), reduce clarity, and lower transcription accuracy. • To avoid those issues, use a single noise-suppression stage: either server-side Krisp on the trunk (krisp_enabled=True) or client/agent-side NC/BVC — not both. If you want help deciding which to use • I can help you choose based on your needs (call quality constraints, STT usage, where you control agents, or latency/bandwidth concerns). • Quick guiding questions: Do you need noise suppression applied before the call reaches your agents (e.g., for telephony-side noise)? Are your agents running their own NC/BVC that you can disable? How important is STT accuracy versus background-noise reduction? If you share your preferred scenario, I’ll recommend the simplest, safest configuration and steps to change it. Sources: Choosing the right noise cancellation for your agent | LiveKit | sip | livekit/livekit_composite | DeepWiki | Python SDKs | livekit/livekit_composite | DeepWiki