We are using a deep neural network for generating raw audio which mimics any human voice and which sounds more natural.