This paper improves speech synthesis using a DGP with SRU for naturalness.
problem Improving naturalness in synthetic speech.
method Deep Gaussian process with a recurrent architecture using SRU.
result SRU-DGP outperforms other models in naturalness of synthetic speech.