Correct, I had loaded randomly the "Italian" voice of the Web Speech API.