---
license: mit
language:
- en
tags:
- voice-conversion
- speech-anonymization
---

<div align="center">
  <img src="figures/logo.png" width="300"/> 
</div>

<p align="center"><strong style="font-size: 22px;">
Self-Supervised LM-Based Zero-Shot Voice Conversion
</strong>
</p>

<p align="center">
♠︎ <a href="https://huggingface.co/ZexinCai/GenVC">Model</a>   | ♣︎ <a href="https://github.com/caizexin/GenVC">Github</a> 
|  ♥︎ <a href="https://arxiv.org/abs/2502.04519">Paper</a> | ♦︎ <a href="https://caizexin.github.io/GenVC/index.html">Demo</a>
</p>

GenVC is an open-source, language model-based zero-shot voice conversion system that leverages self-supervised training and supports streaming voice conversion. **This model card hosts the model checkpoints. For more details on inference and training, please refer to our [GitHub](https://github.com/caizexin/GenVC) repository.**
<!-- --- -->

## Approach
<p align="center">
    <img src="figures/genVC.png" width="100%"/>
</p>


## Features

✅ **Zero-shot Voice Conversion** 

✅ **Streaming VC**

✅ **Self-supervised Training**