Uses WebSocket — Google Speech-to-Text only. Grant microphone when prompted.
Production clients should use /transcript with a Sanctum token. This page uses /transcript-debug for local testing only.
/transcript
/transcript-debug