Users find the voice clone impressive but imperfect; training takes 30-60 min, pacing control is limited, and independent test data is thin.