I have used this repo and got amazing results on my custom datasets (barring a few stuttering issues) that being said, i now have intelligibility thanks this repo, i now plan to train a custom wavenet vocoder, specifically r9y9/wavenet_vocoder, the vocoder works on mel-spectrograms, and i can see that this repo also makes use of mel-spectrograms to generate linear spectrograms, but the spectrograms are of shape x, n_mels * reduction factor . Could anyone please help me out as to how do i extract a melspectrograms of shape x, n_mels so that i can use it within the vocoder, any help would be greatly appreciated.
I have used this repo and got amazing results on my custom datasets (barring a few stuttering issues) that being said, i now have intelligibility thanks this repo, i now plan to train a custom wavenet vocoder, specifically r9y9/wavenet_vocoder, the vocoder works on mel-spectrograms, and i can see that this repo also makes use of mel-spectrograms to generate linear spectrograms, but the spectrograms are of shape
x, n_mels * reduction factor. Could anyone please help me out as to how do i extract a melspectrograms of shape x, n_mels so that i can use it within the vocoder, any help would be greatly appreciated.