Singing Voice Conversion(sing ai cover) 관련 모델 정리
sing ai cover의 포괄적인 분야를 Singing Voice Conversion(SVC) 라고함
1. DiffSVC
https://arxiv.org/abs/2105.13871
DiffSVC: A Diffusion Probabilistic Model for Singing Voice Conversion
Singing voice conversion (SVC) is one promising technique which can enrich the way of human-computer interaction by endowing a computer the ability to produce high-fidelity and expressive singing voice. In this paper, we propose DiffSVC, an SVC system base
arxiv.org
Diffusion 방식을 사용함
2. Retrieval-based Voice Conversion
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI
GitHub - RVC-Project/Retrieval-based-Voice-Conversion-WebUI: Easily train a good VC model with voice data <= 10 mins!
Easily train a good VC model with voice data <= 10 mins! - RVC-Project/Retrieval-based-Voice-Conversion-WebUI
github.com
VITS 기반의 음성 변조를 통한 방식
ov2라는 사전학습모델을 사용하면 20초~1분에 데이터셋만으로도 충분하다.(출처: 나무위키)
3. so-vits-svc
https://github.com/svc-develop-team/so-vits-svc
GitHub - svc-develop-team/so-vits-svc: SoftVC VITS Singing Voice Conversion
SoftVC VITS Singing Voice Conversion. Contribute to svc-develop-team/so-vits-svc development by creating an account on GitHub.
github.com
VITS와 SoftVC를 결합한 방식(https://seastar105.tistory.com/175)
소개한 3개의 모델 중에 RVC가 아직까지도 활성화 돼있는듯