Closed faranaziz closed 3 years ago
The first author is temporarily unavailable due to his mandatory military service (he'll come back on May 20), so I'll answer.
I do not remember such details of our implementation, but either using official/unofficial version and 70-8k/0-8k will be all okay. However, if you're using https://github.com/mindslab-ai/cotatron, make sure that such configurations are identical across acoustic model and the vocoder. In this case, perhaps https://github.com/mindslab-ai/cotatron/issues/14 will be helpful for you.
Thanks you very much.
We used the same mel configuration and mel calculation script with official HiFi-GAN repo. That is, we used the settings of f_min=0 and f_max=8000.
Can you indicate which git repo? Official or unofficial? and the Mel is it still 70-800 or 0-8000?
Thanks you