Inference from OpenUTAU USTx -> DiffSinger DS not Carrying Over Parameters

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

Inference from OpenUTAU USTx -> DiffSinger DS not Carrying Over Parameters #180

Closed sillybillylili closed 3 months ago

sillybillylili commented 3 months ago

Hi! I'm having issues using .ds files generated through OpenUTAU. Parameters such as speaker, speaker mixing, and tension present don't seem to be carried through when exporting to .ds; only pitch is. I'm not sure if this is a DiffSinger or OpenUTAU issue; however, is there a way to preserve these parameters than exporting from OpenUTAU to DiffSinger?

yqzhishen commented 3 months ago

OpenUTAU does not support exporting variance parameters. Actually, it only regards these parameters as intermediate variables and does not persist them at all. You may run inference with your variance checkpoint to obtain the parameters and store them into the ds file, though.