openvpi / DiffSinger

An advanced singing voice synthesis system with high fidelity, expressiveness, controllability and flexibility based on DiffSinger: Singing Voice Synthesis via Shallow Diffusion Mechanism
Apache License 2.0
2.71k stars 285 forks source link

Sample DS File #21

Closed miyu-desu closed 1 year ago

miyu-desu commented 2 years ago

I think it would be good to add the sample DS file to the repo

yqzhishen commented 2 years ago

Sample DS files are added in the samples directory on the refactor branch. We currently use opensvip to obtain DS files, but that tool has only Chinese version. Note that DS file structures may change, and old files may break in the future. A public release containing experimental pretrained models are also provided for usage of different DS files. To use DS files and these models, please read code in main.py carefully. We normally share our updates in communities in China (for example, Bilibili and QQ groups). Sorry for the lack of documentation on GitHub! Feel free to leave more comments if you have questions with DS files, code usage and experimental models.

miyu-desu commented 1 year ago

I am very interested in diffsinger, but I think it is very disappointing to share information in a limited space such as QQ Don't you want to create an information sharing community through GitHub and Discord other than QQ?

yqzhishen commented 1 year ago

Yes, we did realize that we did not share enough information on a global public community. There are some major reasons for this:

But there is exactly something to share:

For international followers of DiffSinger like you, and anyone who would like to provide their help (for example, ideas of improving the network, etc., or GPU resources to help trying out hyper parameters), it is highly recommended to create a topic on GitHub Discussions, as we ourselves will create some in the future.

yqzhishen commented 1 year ago

Close due to inactivity.