yxlllc / HiFiPLN

Multispeaker Community Vocoder Model for DiffSinger

HiFiPLN

Multispeaker Community Vocoder model for DiffSinger

This is the code used to train the "HiFiPLN" vocoder.

Pretrained model is available for download on the official release page.

Why HiFiPLN?

Because a lot of PLN was spent training this thing.

Training

Data preparation

python preproc.py --path dataset --clean --config "configs/hifipln.yaml"

Main model

python train.py --config "configs/hifipln.yaml"

Finetuning

Save the base checkpoint as ckpt/HiFiPLN.cktp then run:

python train.py --config "configs/hifipln-finetune.yaml"

Exporting for use in OpenUtau

python export.py --config configs/hifipln.yaml --output out/hifipln --model CKPT_PATH

Credits

• DiffSinger
• Fish Diffusion
• HiFi-GAN (Paper)
• OpenVPI
• PC-DDSP
• RefineGAN
• UnivNet (Paper)