This repository is a part of our master's course in Speech and Speaker Recognition (DT2119) taken in Spring 2021. In this project, we aim to implement a Voice conversor that only uses an autoencoder, being inspired by this paper . Besides the implementation of the AUTOVC network we also train a speaker encoder from scratch, being inspired by this paper
Voice conversion (VC) is a technique where the speaker characteristics of a source speaker are copied onto the speech contents of a target speaker, transforming the utterance of the target speaker such that it sounds like the source speaker.
Use the package manager pip to install the requirements.
pip install -r requirements.txtThere are two main parts of the training: the training of the speaker encoder and the training of the whole AUTOVC.
- The training and the experiments of the speaker encoder can be found in AUTOVC/speaker_encoder/experiments_speech.ipynb
- The training of the whole AUTOVC can be found in AUTOVC/experiments/test_training.ipynb
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.