Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
A Unified Framework for Expressive Speech Synthesis with Voice Cloning
| Date | Stars |
|---|---|
| 2026-08-03 | 418 |
| 2026-08-13 | 417 |
| 2026-08-21 | 417 |
| 2026-09-07 | 416 |
| 2026-10-07 | 416 |
| 2026-10-08 | 417 |
| 2026-10-16 | 418 |
| 2026-11-04 | 418 |
Today
— stars today
This week
— stars this week
This month
+2 stars this month
Momentum
0.0
growth rate 0.00%/day
<p align="center">
<img src="assets/marco-voice-logo-3.png" alt="Marco-Voice Ecosystem" width="200">
</p>
# Marco-Voice: A Unified Framework for Expressive Speech Synthesis with Voice Cloning
<p align="center">🎧 Empowering Natural Human-Computer Interaction through Expressive Speech Synthesis 🤗</p>
<div align="center">
<img src="https://img.shields.io/badge/License-Apache%202.0-green.svg" alt="License">
<img src="https://img.shields.io/github/issues/AIDC-AI/Marco-Voice?color=red" alt="Open Issues">
<img src="https://img.shields.io/github/issues-closed/AIDC-AI/Marco-Voice?color=green" alt="Closed Issues">
<img src="https://img.shields.io/github/stars/AIDC-AI/Marco-Voice?color=yellow" alt="Stars">
<img src="https://img.shields.io/badge/python-3.10-purple.svg" alt="Python">
<img src="https://img.shields.io/badge/PyTorch-2.0-blue.svg" alt="PyTorch">
<a href="https://huggingface.co/spaces/AIDC-AI/Marco-Voice-TTS"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-yellow" alt="Hugging Face Demo"></a>
<a href="https://huggingface.co/AIDC-AI/Marco-Voice"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Models-orange" alt="Hugging Face Model"></a>
<!-- [](https://arxiv.org/abs/2508.02038) -->
<!-- [](LICENSE) -->
<!-- [](https://huggingface.co/datasets/AIDC-AI/CSEMOTIONS) -->
<!--  -->
<!--  -->
</div>
<p align="center"></p>
<div align="center">
<!-- **Affiliations:** -->
⭐ _**AI Business**_ ⭐
[_**Alibaba International Digital Commerce**_](https://aidc-ai.com)
:octocat: [**Github**](https://github.com/AIDC-AI/Marco-Voice) 🤗 [**Hugging Face**](https://huggingface.co/datasets/AIDC-AI/CSEMOTIONS) 📝 [**Paper**](https://arxiv.org/abs/2508.02038) 🧑💻 [**Model**]() 🗂️ [**Data**](https://github.com/AIDC-AI/Marco-Voice/tree/main/Dataset) 📽️ [**Demo**]()
</div>
## 📖 Overview
<!-- > **Voice Cloning · Emotion Control 😄**
-->
🎯 This project presents a multifunctional speech synthesis system that integrates **voice cloning**, **emotion control**, and **cross-lingual synthesis** within a unified framework. Our goal is to address longstanding challenges in achieving highly expressive, controllable, and natural speech generation that faithfully preserves speaker identity across diverse linguistic and emotional contexts.
<div align="center">
<img src="assets/marco_vocie_fig_v3.jpeg" alt="System Architecture" width="850">
</div>
Our approach introduces an effective speaker-emotion disentanglement mechanism with in-batch contrastive learning, enabling independent manipulation of speaker identity and eemotional style, as well as rotational emotional embedding integration method for smooth emotion control. To support comprehensive training and evaluation, we construct CSEMOTIONS, a high-quality emotional speech dataset containing 10 hours of Mandarin speech from six professional speakers across seven emotional categories. Extensive experiments demonstrate that our system, Marco-Voice, achieves substantial improvements in both objective and subjective metrics. Comprehensive evaluations and analysis were conducted, results show that Marco-Voice delivers competitive performance in terms of speech clarity and emotional richness, representing a substantial advance in the field of expressive neural speech synthesis.
## Audio Samples
Below are sample outputs from our TTS system demonstrating different emotional expressions, with three different speakers:
### Neutral
<table>
<tr>
<td align="center"><b>Speaker001</b></td>
<td align="center"><b>Speaker002</b></td>
<td align="center"><b>Speaker003</b></td>
</tr>
<tr>
<td align="center">
[Speaker001_NExcerpt of 21,409 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:35ee2274ebda0dca, desc:voice cloning, desc:speech synthesis