kittn

/

vocos-mel-48khz-alpha1

like 1

Hello! I have thoroughly enjoyed reading about your research and sincerely appreciate you sharing it.

Using the provided Vocos model checkpoint to convert mel-spectrograms to raw waveforms, unfortunately, resulted in somewhat poor sound quality. I believe this issue could be resolved by matching the 'f_max' setting in config.yaml/feature_extractor/init_arg/ to your settings. Could you please let me know what your f_max setting is?

I would truly appreciate your assistance on this matter. Have a great day!