The algorithm restores the quality of MP3 files compressed to 128 kbps or lower. Model was proposed in this paper and published on github.

The algorithm adds "whispering" effect to vocals. Model was created by SUC-DriverOld. More details here.