So-VITS-SVC

So-VITS-SVC: open-source singing voice conversion

So-VITS-SVC is an open-source project for singing voice conversion. It combines SoftVC-style content representation with VITS-based generation so a trained model can convert a source singing performance into a target voice timbre.

What it is used for

• Research into singing voice conversion and timbre transfer.
• Private experiments with your own licensed or consented voice datasets.
• Vocal demo creation, voice model evaluation, and music AI education.
• Testing preprocessing, pitch extraction, feature encoders, and inference settings.

Important safety and rights notes

Voice conversion can easily cross ethical and legal lines. Do not train on or imitate a person’s voice without permission. Do not publish converted vocals that infringe copyright, performer rights, label rights, or platform policies. Treat So-VITS-SVC as a research tool, not a shortcut for impersonation.

Sources checked

• Official so-vits-svc GitHub repository
• 4.1-Stable README
• Chinese README