data_driven_ai_voice_cloning

Data driven AI voice cloning

This repository is an implementation of the main part of my master thesis in Data science & Engineering. It is divided in two part:

1. Speaker Encoder

1. models: ECAPA-TDNN, wavlm-series
2. data: VoxCeleb1, private dataset

2. Text-to-speech

1. model: FastSpeech2 (microsoft implementation)
2. data: LibriTTS

This two part are then integrated to achieve a Multi Speaker Text to Speech model that is capable of cloning unseen voices starting from about 5 seconds of audio, the ZeroShotFastSpeech2 model.