ContentVec: An Improved Self-Supervised Speech Representation by Disentangling Speakers

This repository provides the official PyTorch implementation of ContentVec.

This is a short video that explains the main concepts of our work. If you find this work useful and use it in your research, please consider citing our paper.

Cite this paper

https://proceedings.mlr.press/v162/qian22b.html

Pre-trained models

The legacy model only contains the representation module, which may be loaded using plain fairseq installation without setting up this code repo.

Load a model

ckpt_path = "/path/to/the/checkpoint_best_legacy.pt"
models, cfg, task = fairseq.checkpoint_utils.load_model_ensemble_and_task([ckpt_path])
model = models[0]

Train a new model

Data preparation

Download the zip file consisting of the following files:

• {train,valid}.tsv waveform list files in metadata
• {train,valid}.km frame-aligned pseudo label files in labels
• dict.km.txt a dummy dictionary in labels
• spk2info.dict a dictionary mapping from speaker id to speaker embedding in metadata

Modify the root directory in the {train,valid}.tsv waveform list files

Setup code repo

Follow steps in setup.sh to setup the code repo

Pretrain ContentVec

Use run_pretrain_single.sh to run on a single node

Use run_pretrain_multi.sh and the corresponding slurm template to run on multiple GPUs and nodes