openvpi / DiffSinger

An advanced singing voice synthesis system with high fidelity, expressiveness, controllability and flexibility based on DiffSinger: Singing Voice Synthesis via Shallow Diffusion Mechanism
Apache License 2.0
2.62k stars 275 forks source link

Using VR model to estimate variance parameters #196

Closed yxlllc closed 2 weeks ago

yxlllc commented 2 weeks ago

Using a harmonic-noise separation model based on VocalRemover to improve the estimation of voicing, breathiness and tension parameters.