Operatic Singing Voice Synthesis Using Diff-SVC
Aoto Sugahara, Soma Kishimoto, Yuji Adachi, Kiyoto Tai, Ryoichi Takashima, Tetsuya Takiguchi · 2023
Singing voice synthesis technology is widely used in the entertainment field, and it has attracted attention as a method for reproducing the singing voices of the deceased or patients who have lost their voices. In recent years, research has also been conducted on synthesizing singing voices with more human-like expression. The purpose of this study is to develop a system that can synthesize operatic singing voices from the voice of a user who has no experience in opera singing. In this paper, we propose a method for synthesizing an operatic singing voice using speechreading, which uses Diff-SVC voice quality conversion to convert the voice quality of opera singing voice to that of the user. To confirm the effectiveness of the proposed method using Diff-SVC, we compare it to a method using CycleGAN-VC2-based voice conversion.