rsxdalv
Ideas for integration
- [x] https://github.com/rakuri255/UltraSinger
- [x] https://github.com/spotify/basic-pitch
- [ ] https://github.com/IAHispano/Applio
- [x] https://github.com/blaisewf/rvc-cli
- [ ] https://github.com/stemrollerapp/stemroller
- [ ] https://github.com/Frikallo/MISST
Vocoders:
- [ ] HiFi-GAN by jik876
- [x] Vocos by gemelo-ai
- [ ] BigVGAN by NVIDIA
- [ ] BigVSAN by sony
- [ ] vocoders by reppy4620
- [ ] vocoder by fishaudio
VC Clients:
- [x] Retrieval-based-Voice-Conversion-WebUI by RVC-Project
- [ ] So-Vits-SVC by svc-develop-team
- [ ] Mangio-RVC-Fork by Mangio621
- [ ] VITS by jaywalnut310
- [ ] Harmonify by Eempostor
- [ ] rvc-trainer by thepowerfuldeez
Pitch Extractors:
- [ ] RMVPE by Dream-High
- [ ] torchfcpe by CNChTu
- [ ] torchcrepe by maxrmorrison
- [ ] anyf0 by SoulMelody
Other:
- [ ] FAIRSEQ by facebookresearch
- [ ] FAISS by facebookresearch
- [ ] ContentVec by auspicious3000
- [ ] audio-slicer by openvpi
- [ ] python-audio-separator by karaokenerds
- [ ] ultimatevocalremovergui by Anjok07
Edited by rsxdalv: turn list into a task list
Thank you!
These I checked and seem to be good:
https://github.com/rakuri255/UltraSinger https://github.com/spotify/basic-pitch
Reasons:
- Good license
- Well known company
- Good chance of future updates
- New feature
- Well written code / easy to integrate
I haven't had the chance to go through all of them, but for example, this one is very-low priority: https://github.com/blaisewf/rvc-cli
Reasons:
- Non-commercial license
- So it cannot be integrated directly
- So people will not make improvements to it
- So it provides a very limited improvement
- The RVC technology is already partially integrated into the UI
- Low popularity - lower popularity projects are less likely to improve the overall adaptation/situation of this tool, unless they are quite good on their own.
@rsxdalv
https://github.com/multimodal-art-projection/YuE https://github.com/deepbeepmeep/YuEGP
MUST SEE
https://github.com/multimodal-art-projection/YuE https://github.com/deepbeepmeep/YuEGP
MUST SEE
@rsxdalv ???
YuE has multiple layers of issues that make it a pain to integrate. For example, they use a modified vocos that obviously interferes with the existing vocos library. Then the installation requires cloning a 1.75gb repository and it's not clear how soon any fixes will be out of date. In reality, quite a few of these projects have massive issues when it comes to software engineering; though some are exceptionally difficult.
Next, YuE does not seem to want to work without flash-attention which is hard to install for Windows users.