Thanks — I’m responding to your question about how to use noise cancellation with a self‑hosted LiveKit server. Short answer • Enhanced agent-side noise cancellation features available in LiveKit Cloud (for example

noise_cancellation.BVC()

/

BVCTelephony()

/ Krisp-style plugins) are not supported in self-hosted deployments. • You can still get noise suppression in a self-hosted setup by handling it yourself — either on the client/device or by integrating third‑party or open‑source models on your servers. What you can do in a self-hosted deployment • Client‑side (recommended first step) ◦ Use the browser/device/WebRTC built‑in features: echo cancellation, automatic gain control, and noise suppression at the capture side. These are lightweight and work without server changes. ◦ If you control native apps, use platform audio APIs or SDKs that offer noise suppression (e.g., OS-level or SDK features). • Replace the Cloud plugin with your own server-side processing ◦ Run an external noise‑suppression model or service (third‑party API or an open‑source model) and process audio streams before forwarding them into LiveKit or to downstream processing. ◦ You can integrate STT/LLM/TTS workflows with custom providers or self‑hosted models in place of LiveKit Cloud plugins. • Remove/disable cloud-only plugin calls ◦ Ensure your agent/server code does not call LiveKit Cloud–specific noise cancellation plugins when running self‑hosted — replace those calls with your chosen alternative. Questions to help me give concrete steps 1. Are you focusing on browser clients, native/mobile clients, or a server‑side processing pipeline? 2. Do you want noise suppression applied on every participant’s microphone before being published, or only on specific agent/consumer audio paths? 3. Would you prefer a managed third‑party API (paid) or an open‑source on‑prem model you run yourself? 4. If server‑side, do you already have a pipeline where audio frames are proxied or processed outside LiveKit (so we can plug in a model)? If you tell me which of the above fits your setup, I’ll give a targeted implementation plan (client config snippets or server pipeline examples and pointers to suitable open‑source models or integration patterns). Sources: Voice AI quickstart | LiveKit Documentation | server-sdk-ruby | livekit/livekit_composite | DeepWiki