Audio leaderboards currently use audio token volume. Speech and transcription need their own unit rollups for seconds and minutes.
