This voicebank is for private use only!
This voicebank will not see a public release as I feel like voice changer programs such as soVITS have a high risk of being used unethically in the wrong hands. As such, there will be no further development of this project is considered dead. Requests can be made for the voice model by emailing me but I will not guarantee that access will be granted!
This project is an AI sing voice changer for Aphelia that uses the same singing data as their NNSVS/ENUNU voicebank.
As soVITS is a voice changer, it works by replacing the audio of an inputted file. As such, this voicebank is technically capable of any language.
Demos
Progress
Currently, the model is trained to 20,000 steps.
| Bank Type | Upload Date | Recording Range | Amount of Data | Training amount | Sample |
|---|---|---|---|---|---|
| soVITS (AI Voice Changer) | PRIVATE | D3-G5 | 4 hours and 11 minutes of English singing 10 minutes of Japanese singing | 20,000 steps |
For a non-voice changer AI voice, please refer to their DiffSinger voicebank!






