Voice enhancement for voice artists using CycleGAN based non-parallel voice conversion

Abstract

This research involves incorporating voice conversion techniques to address the drawbacks identified in the voice casting process. Voice conversion is a technique that retains linguistic information like words, pronunciation, and accent while transforming non-linguistic information like voice quality. The research will conduct experiments to determine the model’s effectiveness on Sinhala language because voice conversion technology has yet to be introduced to the Sinhala language.This study will provide a system that can enhance the voice of a voice artist using state-of-the-art voice conversion models. This will examine how the field of voice conversion can be applied to this problem domain, as well as the most effective voice conversion approaches and a review of existing models. Casting directors and producers will benefit from this system because it will save casting time and expenses.