* Bugfix: code breaks if audio is empty Regression since https://github.com/SYSTRAN/faster-whisper/pull/732 PR