Music FaderNets: Controllable Music Generation Based On High-Level Features via Low-Level Feature Modelling
Audio Examples
Contents
Rhythm Fader Examples
Music snippets are generated by increasing value of zr0 of equal intervals.
Example 1: Original |
Rhythm Level 1 |
Rhythm Level 2 |
|
|
|
|
Rhythm Level 3 |
Rhythm Level 4 |
|
|
|
Example 2: Original |
Rhythm Level 1 |
Rhythm Level 2 |
|
|
|
|
Rhythm Level 3 |
Rhythm Level 4 |
|
|
|
Note Fader Examples
Music snippets are generated by increasing value of zn0 of equal intervals.
Example 1: Original |
Note Level 1 |
Note Level 2 |
|
|
|
|
Note Level 3 |
Note Level 4 |
|
|
|
Example 2: Original |
Note Level 1 |
Note Level 2 |
|
|
|
|
Note Level 3 |
Note Level 4 |
|
|
|
Arousal Style Transfer Examples
Transfer is done by adding a latent shifting vector, lshift = muhigh - mulow (low -> high, and vice versa)
to the latent code before decoding.
ID |
Original |
Low to High Arousal |
H1 |
|
|
H2 |
|
|
H3 |
|
|
H4 |
|
|
H5 |
|
|
H6 |
|
|
H7 |
|
|
H8 |
|
|
H9 |
|
|
H10 |
|
|
ID |
Original |
High to Low Arousal |
L1 |
|
|
L2 |
|
|
L3 |
|
|
L4 |
|
|
L5 |
|
|
L6 |
|
|
L7 |
|
|
L8 |
|
|
L9 |
|
|
L10 |
|
|