The Real Time Voice Cloning (KEY IJP************868)

• M Mohanapriya ,Daram Guru Prasad ,Shashanka K S ,Polakala Somasekhar ,Akshay V Balihallimath

Abstract

A recent study introduces a three-level pipeline for real-time voice cloning, capable of generating natural sounding speech from a few seconds of reference speech without retraining. The framework is based on Google's 2018 paper and aims to be open-sourced. The system can capture and generate audio in real-time and create a practical representation of the voice spoken in a virtual layout. The next step is to train the models with large datasets of tens of hours of audio and observe their strengths and weaknesses. The system can also generate audio from unseen text content, making it a promising tool for text-to-speech applications.