commit
c866e2d8ae
|
@ -44,7 +44,7 @@ The singing voice conversion model uses SoftVC content encoder to extract speech
|
|||
- Added Whisper-PPG encoder support
|
||||
- Added static/dynamic sound fusion
|
||||
- Added loudness embedding
|
||||
- Added feature Retrieve from [RVC](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)
|
||||
- Added Functionality of feature retrieval from [RVC](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)
|
||||
|
||||
### 🆕 Questions about compatibility with the 4.0 model
|
||||
|
||||
|
|
Loading…
Reference in New Issue