MioVocoder is a high-resolution, aliasing-free neural vocoder designed for high-fidelity speech generation. It is a fine-tuned version of the
Pupu-Vocoder (Small) from the
Aliasing-Free Neural Audio Synthesis (AFGen) project.
MioVocoder is specifically optimized to serve as the backend for
MioCodec-25Hz. While the original Pupu-Vocoder is a versatile model, MioVocoder has been fine-tuned with a primary focus on enhancing reconstruction quality for
Japanese speech. By leveraging a large-scale Japanese corpus alongside multilingual data at 44.1kHz, it achieves high robustness and naturalness for various Japanese speaker characteristics.
The model was fine-tuned on a large-scale multilingual corpus, with significant emphasis on Japanese high-fidelity speech data.
As MioVocoder is highly optimized for specific use cases, please note the following:
Since MioVocoder maintains the original Pupu-Vocoder architecture, it can be used with the
official codebase or via the
miocodec helper library.
1from miocodec import load_vocoder
2
3vocoder = load_vocoder(
4 backend="pupu",
5 hf_repo="Aratako/MioVocoder",
6 hf_config_path="config.json",
7 hf_checkpoint_path="model.safetensors",
8).cuda()
If you use MioVocoder in your research, please cite both the original paper and this model checkpoint:
1@article{afgen,
2 title = {Aliasing Free Neural Audio Synthesis},
3 author = {Yicheng Gu and Junan Zhang and Chaoren Wang and Jerry Li and Zhizheng Wu and Lauri Juvela},
4 year = {2025},
5 journal = {arXiv:2512.20211},
6}
1@misc{miovocoder,
2 author = {Chihiro Arata},
3 title = {MioVocoder: High-Resolution Aliasing-Free Neural Vocoder for Japanese Speech},
4 year = {2026},
5 publisher = {Hugging Face},
6 journal = {Hugging Face repository},
7 howpublished = {\url{https://huggingface.co/Aratako/MioVocoder}}
8}
