YingMusic-SVC: Real-World Robust Zero-Shot Singing Voice Conversion with Flow-GRPO and Singing-Specific Inductive Biases

---

---

Overview ✨

Singing voice conversion (SVC) aims to render the target singer’s timbre while preserving melody and lyrics. However, existing zero-shot SVC systems remain fragile in real songs due to harmony interference, F0 errors, and the lack of inductive biases for singing. We propose YingMusic-SVC, a robust zero-shot framework that unifies continuous pre-training, robust supervised fine-tuning, and Flow-GRPO reinforcement learning. Our model introduces a singing-trained RVC timbre shifter for timbre–content disentanglement, an F0-aware timbre adaptor for dynamic vocal expression, and an energy-balanced rectified flow matching loss to enhance high-frequency fidelity. Experiments on a graded multi-track benchmark show that YingMusic-SVC achieves consistent improvements over strong open-source baselines in timbre similarity, intelligibility, and perceptual naturalness—especially under accompanied and harmony-contaminated conditions—demonstrating its effectiveness for real-world SVC deployment.

🔧 Key Features

• Three‑Stage Training Pipeline

  • CPT: Continuous Pre-Training with singing‑trained modules
  • SFT: Robust Supervised Fine-Tuning with F0 perturbation & harmony augmentation
  • RL (Flow‑GRPO): Multi-reward reinforcement learning for perceptual quality

• Singing-Specific Inductive Biases

  • 🎼 RVC-based Timbre Shifter (trained on 120 singers)
  • 🎚️ F0‑Aware Fine-Grained Timbre Adaptor
  • 🔊 Energy-balanced Flow Matching Loss (enhanced high-frequency details)

---

---

News & Updates 🗞️

• 2025-11-26: Released our accompany separator inference CLI and model ckpt
• 2025-11-26: Released gradio app for easy try
• 2025-11-25: Released technical report
• 2025-11-25: Initial YingMusic-SVC inference CLI
• 2025-11-25: Released model checkpoint
• 2025-11-25: Released multi-track benchmark

---

Installation 🛠️

git clone https://github.com/GiantAILab/YingMusic-SVC.git
cd YingMusic-SVC

conda create -n ymsvc python=3.10
conda activate ymsvc
pip install -r requirements.txt

# install ffmpeg & sox
sudo apt update
sudo apt install -y sox libsox-fmt-all
sudo apt install -y ffmpeg

---

Quick Start 🚀

1. accompany separation

cd accom_separation
bash infer.sh

2. SVC Inference

bash my_infer.sh

3. Gradio APP

python gradio_app.py

---

Benchmark Datasets 📚

We provide a graded difficulty benchmark, derived from 100+ multi-track studio songs:

🤗 Download 🔮 Download

---

Pretrained Models 🧪

---

Development Roadmap & TODO 🗺️

• our stem-separator inference CLI & model ckpt
• develop gradio app for YingMusic-SVC
• benchmark one-click eval script

---

Acknowledgements 🙏

This project is built upon:

• Seed-VC

Star 🌟 History

Citation 🧾

If you use YingMusic‑SVC for research, please cite:

@article{chen2025yingmusicsvc,
  title={YingMusic-SVC: Real-World Robust Zero-Shot Singing Voice Conversion with Flow-GRPO and Singing-Specific Inductive Biases},
  author={Chen, Gongyu and Zhang, Xiaoyu and Weng, Zhenqiang and Zheng, Junjie and Shen, Da and Ding, Chaofan and Zhang, Wei-Qiang and Chen, Zihao},
  journal={arXiv preprint arXiv:2512.04793},
  year={2025}
}

---

License 📝

Our code is released under MIT License.