Audio Examples

🎼

Audio companion for "Latent Space Explorations of Singing Voice Synthesis using DDSP" Sound and Music Computing, Aalborg University Code available at https://github.com/juanalonso/DDSP-singing-experiments.

Table of contents

4.1 Pre-evaluation (Violin)

Original audio

Raw resynthesis

Resynthesis with statistical information and masking

4.2.1 Single voice model

Alba

Belen

Eva

Mrallsop

Servando

4.2.2 Multiple voices model - voices2

Common parameters:

-10dB

+0dB

+10dB

4.2.2 Multiple voices model - voices3

Common parameters:

+0 oct, -10dB

+0 oct, +0dB

+0 oct, +10dB

+1 oct, -10dB

+1 oct, +0dB

+1 oct, +10dB

4.3 Single dataset, different spectral parameters

Common parameters:

h:20 n:10

h:20 n:35

h:20 n:65

h:60 n:10

h:60 n:35

h:60 n:65

h:100 n:10

h:100 n:35

h:100 n:60

4.4 Z-encoder

🌈 Over the rainbow

Compare to examples in section 4.2.1

belen, F0 + Amplitude + Z

mrallsop, F0 + Amplitude + Z

felipe_vi, F0 + Amplitude + Z

👃🏼 Smells like teen spirit

belen, F0 + Amplitude + Z

mrallsop, F0 + Amplitude + Z

felipe_vi, F0 + Amplitude + Z

🕺🏻 Never gonna give you up

belen, F0 + Amplitude + Z

mrallsop, F0 + Amplitude + Z

felipe_vi, F0 + Amplitude + Z

5.2 Problems

Silence

Original

Model

Pitch artifacts

Original

Model