hifi-gan
hifi-gan copied to clipboard
What are the correct model changes for sample rate 44100?
Hi, @jik876.
Can you give some advice on how to change the model correctly for the 44100 sample rate? I don't mean hyperparameters in config. For example, how did you choose the kernel size for the convolutions in the discriminators and generator? Maybe it's worth to try doubling their size or increasing convolutions number?
Please understand that we are a bit busy with other work. You can refer #8. Since 44kHz will have longer patterns than 22kHz for same duration, it seems reasonable to enlarge the periods of the discriminator and the receptive fields of the generator, but this should be confirmed through experimentation. I would recommend that you first modify hyperparameters related to upsampling operations in the baseline setting. If the quality is not as good as desired, it will be need to increase other hyperparameters.