Improved noise reduction. The fear will be apprehension mistakes being compounded from speech models.