hey @Plachtaa thanks for this great project, i noticed that the output in svc is in mono format which make the output singing sound flat and like suffocating and not dynamic and rich like if its stereo, i fixed that problem by using plugin called "ozone imager 2" for free in adobe audition and its true that it conver the sound from mono to stereo but in cost of loosing some quality not like the raw original
and also please update the 0F condition to improved pitch accuracy and more expressive output and reduced artifacts, there are many new libraries that can be used like (PyWorld, librosa, torchcrepe, ESPnet, S3PRL, conditional-method)
and @Plachtaa how are you doing? its been several months without you being active i hope you are okay