search

maum-ai / nuwave2

NU-Wave 2: A General Neural Audio Upsampling Model for Various Sampling Rates @ INTERSPEECH 2022
https://mindslab-ai.github.io/nuwave2
BSD 3-Clause "New" or "Revised" License
278 stars 21 forks source link

Just wondering if the pretrained model will produce good results for non-English voice. #3

Closed Michaelwhite34 closed 2 years ago

Seungwoo0326 commented 2 years ago

Hi! Thank you for the great question.

I checked whether NU-Wave 2 model can produce good results for Korean and here is an example of the KSS dataset (22050 Hz -> 48000 Hz). I think the result is quite good. It seems to work well because the audio super-resolution task does not depend on the language, unlike other tasks like TTS or STT.

I hope this helps!

wav_l result result_kss.zip