facebookresearch / speech-resynthesis

An official reimplementation of the method described in the INTERSPEECH 2021 paper - Speech Resynthesis from Discrete Disentangled Self-Supervised Representations.
Other
390 stars 54 forks source link

Upsampling Hubert features from 50hz to 100 hz. #8

Closed shangeth closed 2 years ago

shangeth commented 2 years ago

The output of Hubert features is 50 Hz(ie: 1s audio gives 50 timestep/frame), how are you upsampling it to 100Hz?