Abstract: End-to-end (E2E) sequence-to-sequence (S2S) neural text-to-speech (TTS) models and E2E-S2S neural voice conversion (VC) models can achieve high-quality speech synthesis with a single neural ...