Audio samples from "Disentangling speech from surroundings in a neural audio codec"
Authors: Ahmed Omran, Neil Zeghidour, Zalán Borsos,
Félix de Chaumont Quitry, Malcolm Slaney,
Marco Tagliasacchi
The audio examples on this page were randomly selected from evaluation splits
of the datasets. For noisy samples, neither the speech nor noise components
were seen during training. For examples using reverberant speech, the room
impulse responses were also drawn from an evaluation set withheld during
training.
Disentangling speech from noise
Input sample
Reconstruction
Decoding speech partition
Decoding noise partition
Example 1
Example 2
Example 3
Example 4
Scaling the noise partition by a weight factor
1
0.75
0.5
0.25
0
Example 1
Example 2
Example 3
Example 4
Swapping noise between samples
Input A
Input B
Output A
Output B
Example 1
Example 2
Example 3
Example 4
Disentangling speech from reverberation
Input sample
Reconstruction
Decoding speech partition
Example 1
Example 2
Example 3
Example 4
Scaling the reverberation partition by a weight factor