I would excitement with different transcription models to see if you get a better outcome