Neural Analysis and Synthesis: Reconstructing Speech from Self-Supervised Representations