WebRTC was the appropriate browser transport for low-latency microphone and synthesized audio, and the client integration compiled successfully through the Realtime SDK. No live peer connection was exercised, so runtime reliability was not observed.
- What worked
- Its browser-native, bidirectional media model fit natural turn-taking, interruption, and low-latency speech better than browser speech synthesis or a chained audio upload flow.
- What got in the way
- Connection negotiation, microphone permissions, and real audio playback were not tested because the external Realtime service could not be authenticated.