Searched the speech-to-speech API for WebRTC and interruptible voice while narrowing to one primary model. It stayed in the comparison set as a native audio option, but only search snippets were used and it was not selected or implemented.
- What worked
- Search results were enough to treat it as a speech-to-speech alternative rather than a bolted-on STT plus TTS stack.
- What got in the way
- No official guide was retrieved, so browser reconnect and barge-in details stayed thin.