Project
so-vits-svc
svc-develop-team/so-vits-svc · Python · AGPL-3.0
Classic open singing voice conversion stack
so-vits-svc powered a wave of AI covers. Still relevant when you need SVC-specific workflows alongside RVC.
12.9 / 100 · Score
★ 28k Stars
+-8 30-day growth
40 min Time to first run
Not required API key
Score
Trend 14
Activity 8
Quality 70
Community 63
Relevance 100
Setup at a glance
DeployCli, Web
GPU6 GB
API keyNot required
UI languagesen, zh
Doc languageszh, en
Install
git clone https://github.com/svc-develop-team/so-vits-svc.git Deep dive
so-vits-svc is the classic open singing-voice-conversion stack: it re-sings an existing vocal in another timbre while preserving the melody — the workhorse behind a huge share of AI cover songs.
The current iteration takes its feature input from the 12th layer of Content Vec, adds shallow-diffusion for quality, and supports Whisper-PPG as an encoder option, each documented in the repo.
It sits in the same family tree as DiffSinger (both from the openvpi lineage), and pretrained resources are distributed via the VoiceConversionWebUI project.
Use cases
Singing conversion
Character song covers
SVC research
Good for / Not ideal for
Good for
Cover creators familiar with SVC
Not ideal for
Absolute beginners — start with RVC WebUI