ShiroRVC Resources
Pretrained models and other files downloaded by ShiroRVC.
License
The files in this repository are for non-commercial use only, under CC BY-NC-SA 4.0 (https://creativecommons.org/licenses/by-nc-sa/4.0/). This also applies to any model fine-tuned from them.
Rectified pretrains
Rectified_pretrains/ (the flow pretrain) and the PCPH-BigVGAN vocoder
pretrain were trained on:
- EARS (CC BY-NC 4.0), https://github.com/facebookresearch/ears_dataset Richter et al., "EARS: An Anechoic Fullband Speech Dataset Benchmarked for Speech Enhancement and Dereverberation", Interspeech 2024.
- M4Singer (CC BY-NC-SA 4.0), https://github.com/M4Singer/M4Singer Zhang et al., "M4Singer: A Multi-Style, Multi-Singer and Musical Score Provided Mandarin Singing Corpus", NeurIPS 2022 Datasets and Benchmarks.
- GTSinger (CC BY-NC-SA 4.0), https://github.com/AaronZ345/GTSinger Zhang et al., "GTSinger: A Global Multi-Technique Singing Corpus with Realistic Music Scores for All Singing Tasks", NeurIPS 2024 Datasets and Benchmarks.
OpenVPI vocoder
vocoders/pc_nsf_hifigan_44.1k_hop512_128bin_vocoder.pth is a converted copy
of the pc-nsf-hifigan vocoder by Team OpenVPI
(https://github.com/openvpi/SingingVocoders), licensed CC BY-NC-SA 4.0. It was
converted to this project's export format; the weights are unchanged.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support