xiph
cepstrum VQ train
When I change the dataset and wanna train a new model, should I change the ceps_codebooks.c at first？
I didn't change the ceps_bodebooks.c at all, and only got nnet_data.* from the lpcnet*.h5 file.
I doubt this is why the final output voice was blurry and fuzzy.
If so, how should I use ceps_vq_train.c to get my own codebooks ?
When I change the dataset and wanna train a new model, should I change the ceps_codebooks.c at first？
I didn't change the ceps_bodebooks.c at all, and only got nnet_data.* from the lpcnet*.h5 file.
I doubt this is why the final output voice was blurry and fuzzy.
If so, how should I use ceps_vq_train.c to get my own codebooks ?
Brother, have you solved this problem? I am also converting the sampling rate to be processed. I want to switch to 8KHz to process speech, but I don't know how to get my experience with codebooks
