---
title: OmniVoice-Emilia
canonical_url: "https://www.modelscope.ai/models/k2-fsa/OmniVoice-Emilia"
md_url: "https://www.modelscope.ai/models/k2-fsa/OmniVoice-Emilia.md"
repository: k2-fsa/OmniVoice-Emilia
last_updated: 2026-06-05
pipeline_tag: text-to-speech
tasks:
  - text-to-speech
model_type:
  - omnivoice
architectures:
  - OmniVoice
parameters: 612.6M
tensor_type:
  - F32
  - I64
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 163
stars: 0
tags:
  - zero-shot
  - multilingual
  - voice-cloning
---

# OmniVoice-Emilia

> OmniVoice-Emilia - An open-source model by k2-fsa on ModelScope.        

- **Repository**: k2-fsa/OmniVoice-Emilia
- **Tasks**: text-to-speech
- **Parameters**: 612.6M
- **Tags**: zero-shot, multilingual, voice-cloning
- **Downloads**: 163
- **Stars**: 0
- **Last updated**: 2026-06-05

Source: https://www.modelscope.ai/models/k2-fsa/OmniVoice-Emilia

---

# OmniVoice 🌍

<p align="center">
  <img width="200" height="200" alt="OmniVoice" src="https://zhu-han.github.io/omnivoice/pics/omnivoice.jpg" />
</p>

<p align="center">
  <a href="https://huggingface.co/k2-fsa/OmniVoice"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Model-FFD21E" alt="Hugging Face Model"></a>
  &nbsp;
  <a href="https://huggingface.co/spaces/k2-fsa/OmniVoice"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Space-blue" alt="Hugging Face Space"></a>
  &nbsp;
  <a href="https://huggingface.co/papers/2604.00688"><img src="https://img.shields.io/badge/arXiv-Paper-B31B1B.svg"></a>
  &nbsp;
  <a href="https://github.com/k2-fsa/OmniVoice"><img src="https://img.shields.io/badge/GitHub-Code-181717?logo=GitHub" alt="GitHub Code"></a>
  &nbsp;
  <a href="https://zhu-han.github.io/omnivoice"><img src="https://img.shields.io/badge/GitHub.io-Demo_Page-blue?logo=GitHub&style=flat-square"></a>
</p>


This OmniVoice variant was trained exclusively on the Chinese and English subsets of the Emilia dataset and corresponds to the "OmniVoice-Emilia" model described in our paper. It is intended for researchers aiming to reproduce the experimental results reported therein. For regular end users seeking superior performance, we recommend using the full-dataset-trained OmniVoice checkpoint [OmniVoice](https://huggingface.co/k2-fsa/OmniVoice) instead.


When using this checkpoint, set `denoise = False` and `lang_id = None`: the model was trained without prompt denoising or language-ID conditioning.

## Citation

```bibtex
@article{zhu2026omnivoice,
      title={OmniVoice: Towards Omnilingual Zero-Shot Text-to-Speech with Diffusion Language Models},
      author={Zhu, Han and Ye, Lingxuan and Kang, Wei and Yao, Zengwei and Guo, Liyong and Kuang, Fangjun and Han, Zhifeng and Zhuang, Weiji and Lin, Long and Povey, Daniel},
      journal={arXiv preprint arXiv:2604.00688},
      year={2026}
}
```