Short answer: use the WebRTC connection for your Agent session. WebRTC adds built‑in acoustic echo cancellation so the agent’s speaker output won’t re‑trigger the mic, and you can keep “interruptions” enabled.
What to do
Start the session in WebRTC mode instead of raw WebSocket:
React SDK: call startSession with connectionType: "webrtc". WebRTC is explicitly supported and is the recommended path for real‑time voice. (elevenlabs.io)
If you prefer a prebuilt UI, the Conversation Bar component also uses WebRTC under the hood. (ui.elevenlabs.io)
Keep interruptions on: in your agent settings, enable Interruptions under Advanced → Client Events. This controls barge‑in behavior and is independent of the transport you use. (elevenlabs.io)
Why this works
ElevenLabs’ WebRTC support “unlocks best‑in‑class echo cancellation and background noise removal,” which addresses speaker‑to‑mic self‑trigger without muting the mic. (elevenlabs.io)
If you’re using the React SDK, you can dynamically lower the agent’s output volume when the agent is speaking (e.g., via onModeChange and the controlled volume state) to further reduce pickup, while leaving the mic live. (elevenlabs.io)
Note
The documentation highlights echo cancellation as a benefit of the WebRTC transport; it does not document equivalent echo‑cancellation for the raw WebSocket path. If you remain on WebSockets, the recommended fix in the docs is to switch to WebRTC for this capability. (elevenlabs.io)
fiery gull
median zealot