search

kan-bayashi / ParallelWaveGAN

Unofficial Parallel WaveGAN (+ MelGAN & Multi-band MelGAN & HiFi-GAN & StyleMelGAN) with Pytorch
https://kan-bayashi.github.io/ParallelWaveGAN/
MIT License
1.55k stars 340 forks source link

Questions on SVS #399

Open ghost opened 1 year ago

ghost commented 1 year ago

Hi, im highly interested singing voice synthesis, i saw that there are pretrained models for svs but i dont understand how to specify what the model should sing eg is it possible to have it sing a text?

ftshijt commented 1 year ago

Thanks for your interest on singing voice synthesis. This pre-trained model in this repo is for the vocoder training (i.e., converting acoustic features into waveform) instead of acoustic modeling. If you are interested in acoustic modeling (e.g., score2wav, score2feat), please feel free to check either ESPnet or Muskits, where we do offer acoustic modeling training with vocoders provided from this repo.

ghost commented 1 year ago

Hi thank you for your time and the information provided, ill make sure to check it out. :)

lhc991025 commented 4 months ago

Thanks for your interest on singing voice synthesis. This pre-trained model in this repo is for the vocoder training (i.e., converting acoustic features into waveform) instead of acoustic modeling. If you are interested in acoustic modeling (e.g., score2wav, score2feat), please feel free to check either ESPnet or Muskits, where we do offer acoustic modeling training with vocoders provided from this repo.

I want to train a vocoder with my own data. How do I do that?

ftshijt commented 4 months ago

You can follow any of the existing recipe, but replacing the wav.scp with your own data.