I imagine the whisper transcript has less errors.