So-VITS-SVC

So-VITS-SVC is an open-source project that has spawned many, many forks. It is a Singing Voice Conversion model that uses SoftVC content encoder to extract speech features from source audio.

The feature vectors are fed into VITS without the need for conversion to a text-based intermediate representation.

Therefore, the pitch and intonation of the original audio are preserved.

Meanwhile, the vocoder was replaced with NSF HiFi-GAN to solve the problem of sound interruption.