Real-Time Voice Cloning Implementation Using SV2TTS Deep Learning Framework

Summary

This repository implements a real-time voice cloning system called SV2TTS (Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis) that can clone a voice from just 5 seconds of audio and generate arbitrary speech in real-time. The project was developed as a master's thesis and uses a three-stage deep learning framework: first creating a digital voice representation from audio, then using that representation to generate speech from text input with a real-time vocoder.