Author: Gwendal Le Vaillant (ISIA Lab, University of Mons)
This page provides complementary and detailed results comparing three methods:
- two preset interpolation techniques: the linear parametric interpolation, and interpolation using SPINVAE
- state-of-the-art sound morphing using SMT (Sound Morphing Toobox)
For each preset interpolation or morphing example, trajectories of key timbre features are displayed. To maintain clarity, only four features (those most closely correlated with subjective evaluations of morphing quality) are shown.
The following additional examples are available:
- Example 6: from “E.Piano 23” to “B3 Organ 3”
- Example 7: from “AnlgSyn.45” to “ClinkieBel”
- Example 8: from “LOG DRUMS” to “Hard.Money”
- Example 9: from “CHIMES” to “FUNKEYS”
Interpolation example 6
|
Start preset, step 1/9: "E.Piano 23" End preset, step 9/9: "B3 Organ 3" |
|||||||||
|---|---|---|---|---|---|---|---|---|---|
| Step 1/9 | Step 2/9 | Step 3/9 | Step 4/9 | Step 5/9 | Step 6/9 | Step 7/9 | Step 8/9 | Step 9/9 | |
| Linear parametric preset interpolation (linearity = -0.98 ; smoothness = -60.5) | |||||||||
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
| SPINVAE preset interpolation (linearity = -0.39 ; smoothness = -39.2) | |||||||||
| Start sound reconstruction: MFCCD = 0 PEMO-Q ODG = 0 |
End sound reconstruction: MFCCD = 0.4 PEMO-Q ODG = -0.01 |
||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
| SMT sound morphing (linearity = -0.25 ; smoothness = -11.5) | |||||||||
| Start sound reconstruction: MFCCD = 3.9 PEMO-Q ODG = -0.68 |
End sound reconstruction: MFCCD = 5.5 PEMO-Q ODG = -1.66 |
||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
Interpolation example 7
|
Start preset, step 1/9: "AnlgSyn.45" End preset, step 9/9: "ClinkieBel" |
|||||||||
|---|---|---|---|---|---|---|---|---|---|
| Step 1/9 | Step 2/9 | Step 3/9 | Step 4/9 | Step 5/9 | Step 6/9 | Step 7/9 | Step 8/9 | Step 9/9 | |
| Linear parametric preset interpolation (linearity = -0.55 ; smoothness = -64.4) | |||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
| SPINVAE preset interpolation (linearity = -0.42 ; smoothness = -41.8) | |||||||||
| Start sound reconstruction: MFCCD = 0.5 PEMO-Q ODG = -0.03 |
End sound reconstruction: MFCCD = 0.1 PEMO-Q ODG = -0.01 |
||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
| SMT sound morphing (linearity = -0.30 ; smoothness = -15.8) | |||||||||
| Start sound reconstruction: MFCCD = 4.2 PEMO-Q ODG = -0.67 |
End sound reconstruction: MFCCD = 6.5 PEMO-Q ODG = -0.49 |
||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
Interpolation example 8
|
Start preset, step 1/9: "LOG DRUMS" End preset, step 9/9: "Hard.Money" |
|||||||||
|---|---|---|---|---|---|---|---|---|---|
| Step 1/9 | Step 2/9 | Step 3/9 | Step 4/9 | Step 5/9 | Step 6/9 | Step 7/9 | Step 8/9 | Step 9/9 | |
| Linear parametric preset interpolation (linearity = -0.82 ; smoothness = -126.4) | |||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
| SPINVAE preset interpolation (linearity = -0.37 ; smoothness = -29.5) | |||||||||
| Start sound reconstruction: MFCCD = 0.0 PEMO-Q ODG = -0.00 |
End sound reconstruction: MFCCD = 0.3 PEMO-Q ODG = -0.00 |
||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
| SMT sound morphing (linearity = -0.50 ; smoothness = -21.5) | |||||||||
| Start sound reconstruction: MFCCD = 5.5 PEMO-Q ODG = -2.27 |
End sound reconstruction: MFCCD = 10.6 PEMO-Q ODG = -0.93 |
||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
Interpolation example 9
|
Start preset, step 1/9: "CHIMES" End preset, step 9/9: "FUNKEYS" |
|||||||||
|---|---|---|---|---|---|---|---|---|---|
| Step 1/9 | Step 2/9 | Step 3/9 | Step 4/9 | Step 5/9 | Step 6/9 | Step 7/9 | Step 8/9 | Step 9/9 | |
| Linear parametric preset interpolation (linearity = -0.44 ; smoothness = -39.8) | |||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
| SPINVAE preset interpolation (linearity = -0.37 ; smoothness = -38.6) | |||||||||
| Start sound reconstruction: MFCCD = 2.7 PEMO-Q ODG = -2.55 |
End sound reconstruction: MFCCD = 0.6 PEMO-Q ODG = -0.05 |
||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||
| SMT sound morphing (linearity = -0.49 ; smoothness = -20.3) | |||||||||
| Start sound reconstruction: MFCCD = 42.3 PEMO-Q ODG = -2.43 |
End sound reconstruction: MFCCD = 9.0 PEMO-Q ODG = -0.54 |
||||||||
|
|
|
|
|
|
|
|
|
|
|
 |
|||||||||