https://github.com/hrnoh/f0-autovc

Pytorch implementation of "f0-consistent many-to-many non-parallel voice conversion via conditional autoencoder"
https://github.com/hrnoh/f0-autovc

Last synced: 3 days ago
JSON representation

Pytorch implementation of "f0-consistent many-to-many non-parallel voice conversion via conditional autoencoder"

• Host: GitHub
• URL: https://github.com/hrnoh/f0-autovc
• Owner: hrnoh
• Created: 2020-10-13T11:30:15.000Z (about 4 years ago)
• Default Branch: main
• Last Pushed: 2020-11-06T08:49:15.000Z (about 4 years ago)
• Last Synced: 2024-08-02T13:28:27.951Z (3 months ago)
• Language: Python
• Size: 9.11 MB
• Stars: 28
• Watchers: 2
• Forks: 4
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-voice-conversion - [f0-autovc - WavRNN]](https://github.com/freenowill/AutoVC-WavRNN), [[AutoVC]](https://github.com/CODEJIN/AutoVC) (Overview / Voice Conversion)

README

        
## F0-AUTOVC: F0-Consistent Many-to-Many Non-Parallel Voice Conversion via Conditional Autoencoder  

This repository provides a PyTorch implementation of the paper [F0-AUTOVC](https://arxiv.org/abs/2004.07370).

Based on

- https://github.com/auspicious3000/autovc

- https://github.com/auspicious3000/SpeechSplit

- https://github.com/christopher-beckham/amr

## Dependencies

- Python 3.7

- Pytorch 1.6.0

- TensorFlow

- Numpy

- librosa

- tqdm

## Usage

1. Prepare dataset


    we used the [VCTK dataset](http://www.udialogue.org/download/cstr-vctk-corpus.html) as used in original paper.  

    But, you can use your own dataset.

    

2. Prepare the speaker to gender file as shown in nikl_spk.txt and run ```make_spk2gen.py```  

    * Format  

    speaker1 gender1  

    speaker2 gender2  

    

    * Example:  

        p225    W  

        p226    M  

        p301    W  

        p302    W  

        .  

        .

3. Preprocess data using ```preprocess.py```  

4. Run ```task_launcher.py```