Five-minute quick start — open a session, send text, hear the AI read it.
End-to-end WebRTC voice call — start, mid-call read-aloud, DTMF, hang up.