Sound demos for "APNet: An All-Frame-Level Neural Vocoder Incorporating Direct Prediction of Amplitude and Phase Spectra"

 

 

Comparsion among vocoders:

Analysis-Synthesis Task

Example 1
Natural          
       
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET HiNet MB-MelGAN
Example 2
Natural          
       
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET HiNet MB-MelGAN
Example 3
Natural          
       
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET HiNet MB-MelGAN
Example 4
Natural          
       
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET HiNet MB-MelGAN
Example 5
Natural          
       
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET HiNet MB-MelGAN

TTS Task

Example 1
Natural        
     
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET MB-MelGAN
Example 2
Natural        
     
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET MB-MelGAN
Example 3
Natural        
     
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET MB-MelGAN
Example 4
Natural        
     
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET MB-MelGAN
Example 5
Natural        
     
APNet HiFi-GAN v1 HiFi-GAN v2 ISTFTNET MB-MelGAN

Ablation studies:

(1) APNet vs APNet wo PPEA :
  APNet APNet wo PPEA
Example 1
Example 2
Example 3
Example 4
Example 5
(2) APNet vs APNet wo A :
  APNet APNet wo A
Example 1
Example 2
Example 3
Example 4
Example 5
(3) APNet vs APNet wo IP :
  APNet APNet wo P
Example 1
Example 2
Example 3
Example 4
Example 5
(4) APNet vs APNet wo GD :
  APNet APNet wo GD
Example 1
Example 2
Example 3
Example 4
Example 5
(5) APNet vs APNet wo PTD :
  APNet APNet wo PTD
Example 1
Example 2
Example 3
Example 4
Example 5
(6) APNet vs APNet wo P :
  APNet APNet wo P/GD/PTD
Example 1
Example 2
Example 3
Example 4
Example 5
(7) APNet vs APNet wo C :
  APNet APNet wo C
Example 1
Example 2
Example 3
Example 4
Example 5
(8) APNet vs APNet wo RI :
  APNet APNet wo RI
Example 1
Example 2
Example 3
Example 4
Example 5
(9) APNet vs APNet wo W :
  APNet APNet wo W
Example 1
Example 2
Example 3
Example 4
Example 5