SALTTS: Leveraging Self-Supervised Speech Representations for improved Text-to-Speech Synthesis