Diff-Pitcher: Diffusion-based Singing Voice Pitch Correction

Official Pytorch Implementation of Diff-Pitcher

🤗 Download model form huggingface

🎵 Listen to examples

---

Thank you all for your interest in this research project. I am currently optimizing the model's performance and computation efficiency.

A new version of Diff-Pitcher is in development and will feature a user-friendly Hugging Face space.

---

Diff-Pitcher

• Todo List
• Code Examples
• References
• Acknowledgement

Todo

• Update codes and demo
• Support 🤗 Diffusers
• Upload checkpoints
• Pipeline tutorial
• Diff-Pitcher v2 with Huggingface space demo

Examples

• Download checkpoints: 🎒ckpts
• Prepare environment: requirements.txt
• Feel free to try:
  • template-based automatic pitch correction: template_based_apc.py
  • score-based automatic pitch correction: score_based_apc.py

References

If you find the code useful for your research, please consider citing:

@inproceedings{hai2023diff,
  title={Diff-Pitcher: Diffusion-Based Singing Voice Pitch Correction},
  author={Hai, Jiarui and Elhilali, Mounya},
  booktitle={2023 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA)},
  pages={1--5},
  year={2023},
  organization={IEEE}
}

This repo is inspired by:

@article{popov2021diffusion,
  title={Diffusion-based voice conversion with fast maximum likelihood sampling scheme},
  author={Popov, Vadim and Vovk, Ivan and Gogoryan, Vladimir and Sadekova, Tasnima and Kudinov, Mikhail and Wei, Jiansheng},
  journal={arXiv preprint arXiv:2109.13821},
  year={2021}
}

@inproceedings{liu2022diffsinger,
  title={Diffsinger: Singing voice synthesis via shallow diffusion mechanism},
  author={Liu, Jinglin and Li, Chengxi and Ren, Yi and Chen, Feiyang and Zhao, Zhou},
  booktitle={Proceedings of the AAAI conference on artificial intelligence},
  volume={36},
  number={10},
  pages={11020--11028},
  year={2022}
}

Acknowledgement

Welcome to LCAP! < LCAP (jhu.edu)

We borrow code from following repos:

• Diffusion Schedulers are based on 🤗 Diffusers
• 2D UNet is based on DiffVC