def commit_user_turn(
        self,
        *,
        audio_detached: bool,
        transcript_timeout: float,
        stt_flush_duration: float = 2.0,
    ) -> None:
        if not self._stt or self._closing.is_set():
            return

        async def _commit_user_turn() -> None:
            if time.time() - self._last_final_transcript_time > 0.5:
                # if the last final transcript is received more than 0.5s ago
                # append a silence frame to the stt to flush the buffer

                self._final_transcript_received.clear()

does this essentially mean that if we're doing a push to talk sort of thing, and the person says a statement, waits a bit, the STT emits a speech final, and then it takes them >0.5s to "submit" what they said -> then we're expecting the STT to emit another speech final event and are willing to wait

transcript_timeout

long for it? Do many stt providers support this? I don't see why they'd send another speech final event just because

stt_flush_duration

seconds of silence was sent?