Is it possible for the user to record audio directly and then send this audio to OpenAI for transcription?