FFTimbre Showcase
Audio samples and matching plots
Bowed Cello
Legend
- FFT frame (bowed cello)
- Target spectra
- Optimized FM — DE + cosine
- Optimized FM — DE + euclidean
- Optimized FM — DE + itakura_saito
- Optimized FM — DE + kl
- Optimized FM — DE + manhattan
- Optimized FM — DE + mfcc
- Optimized FM — DE + pearson
- Optimized FM — DE + spectral_convergence
- Optimized FM — DA + cosine
- Optimized FM — DA + euclidean
- Optimized FM — DA + itakura_saito
- Optimized FM — DA + kl
- Optimized FM — DA + manhattan
- Optimized FM — DA + mfcc
- Optimized FM — DA + pearson
- Optimized FM — DA + spectral_convergence
- Optimized FM — BH + cosine
- Optimized FM — BH + euclidean
- Optimized FM — BH + itakura_saito
- Optimized FM — BH + kl
- Optimized FM — BH + manhattan
- Optimized FM — BH + mfcc
- Optimized FM — BH + pearson
- Optimized FM — BH + spectral_convergence
- Optimized Additive — DE + cosine
- Optimized Additive — DE + euclidean
- Optimized Additive — DE + itakura_saito
- Optimized Additive — DE + kl
- Optimized Additive — DE + manhattan
- Optimized Additive — DE + mfcc
- Optimized Additive — DE + pearson
- Optimized Additive — DE + spectral_convergence
- Optimized Additive — DA + cosine
- Optimized Additive — DA + euclidean
- Optimized Additive — DA + itakura_saito
- Optimized Additive — DA + kl
- Optimized Additive — DA + manhattan
- Optimized Additive — DA + mfcc
- Optimized Additive — DA + pearson
- Optimized Additive — DA + spectral_convergence
- Optimized Additive — BH + cosine
- Optimized Additive — BH + euclidean
- Optimized Additive — BH + itakura_saito
- Optimized Additive — BH + kl
- Optimized Additive — BH + manhattan
- Optimized Additive — BH + mfcc
- Optimized Additive — BH + pearson
- Optimized Additive — BH + spectral_convergence
FFT Frame based on bowed cello sound
Source: tsv/cello_single.tsv
Target spectra
Reference spectrogram and waveform for the bowed cello FFT frame.
Results below are from a batch run (maxiter=500) across metrics using Differential Evolution.
Optimized FM with DE + cosine
FM resynthesis optimized with Differential Evolution using cosine similarity.
Optimized FM with DE + euclidean
FM resynthesis optimized with Differential Evolution using Euclidean distance.
Optimized FM with DE + itakura_saito
FM resynthesis optimized with Differential Evolution using Itakura–Saito divergence.
Optimized FM with DE + kl
FM resynthesis optimized with Differential Evolution using KL divergence.
Optimized FM with DE + manhattan
FM resynthesis optimized with Differential Evolution using Manhattan distance.
Optimized FM with DE + mfcc
FM resynthesis optimized with Differential Evolution using MFCC distance.
Optimized FM with DE + pearson
FM resynthesis optimized with Differential Evolution using Pearson correlation.
Optimized FM with DE + spectral_convergence
FM resynthesis optimized with Differential Evolution using spectral convergence.
Results below are from a batch run (maxiter=500) across metrics using Dual Annealing.
Optimized FM with DA + cosine
FM resynthesis optimized with Dual Annealing using cosine similarity.
Optimized FM with DA + euclidean
FM resynthesis optimized with Dual Annealing using Euclidean distance.
Optimized FM with DA + itakura_saito
FM resynthesis optimized with Dual Annealing using Itakura–Saito divergence.
Optimized FM with DA + kl
FM resynthesis optimized with Dual Annealing using KL divergence.
Optimized FM with DA + manhattan
FM resynthesis optimized with Dual Annealing using Manhattan distance.
Optimized FM with DA + mfcc
FM resynthesis optimized with Dual Annealing using MFCC distance.
Optimized FM with DA + pearson
FM resynthesis optimized with Dual Annealing using Pearson correlation.
Optimized FM with DA + spectral_convergence
FM resynthesis optimized with Dual Annealing using spectral convergence.
Results below are from a batch run (maxiter=500) across metrics using Basin Hopping.
Optimized FM with BH + cosine
FM resynthesis optimized with Basin Hopping using cosine similarity.
Optimized FM with BH + euclidean
FM resynthesis optimized with Basin Hopping using Euclidean distance.
Optimized FM with BH + itakura_saito
FM resynthesis optimized with Basin Hopping using Itakura–Saito divergence.
Optimized FM with BH + kl
FM resynthesis optimized with Basin Hopping using KL divergence.
Optimized FM with BH + manhattan
FM resynthesis optimized with Basin Hopping using Manhattan distance.
Optimized FM with BH + mfcc
FM resynthesis optimized with Basin Hopping using MFCC distance.
Optimized FM with BH + pearson
FM resynthesis optimized with Basin Hopping using Pearson correlation.
Optimized FM with BH + spectral_convergence
FM resynthesis optimized with Basin Hopping using spectral convergence.
Results below are from a batch run (maxiter=500) across metrics using Differential Evolution.
Optimized Additive with DE + cosine
Additive resynthesis optimized with Differential Evolution using cosine similarity.
Optimized Additive with DE + euclidean
Additive resynthesis optimized with Differential Evolution using Euclidean distance.
Optimized Additive with DE + itakura_saito
Additive resynthesis optimized with Differential Evolution using Itakura–Saito divergence.
Optimized Additive with DE + kl
Additive resynthesis optimized with Differential Evolution using KL divergence.
Optimized Additive with DE + manhattan
Additive resynthesis optimized with Differential Evolution using Manhattan distance.
Optimized Additive with DE + mfcc
Additive resynthesis optimized with Differential Evolution using MFCC distance.
Optimized Additive with DE + pearson
Alternate run of Additive resynthesis with Differential Evolution using Pearson correlation.
Optimized Additive with DE + spectral_convergence
Additive resynthesis optimized with Differential Evolution using spectral convergence.
Results below are from a batch run (maxiter=500) across metrics using Dual Annealing.
Optimized Additive with DA + cosine
Additive resynthesis optimized with Dual Annealing using cosine similarity.
Optimized Additive with DA + euclidean
Additive resynthesis optimized with Dual Annealing using Euclidean distance.
Optimized Additive with DA + itakura_saito
Additive resynthesis optimized with Dual Annealing using Itakura–Saito divergence.
Optimized Additive with DA + kl
Additive resynthesis optimized with Dual Annealing using KL divergence.
Optimized Additive with DA + manhattan
Additive resynthesis optimized with Dual Annealing using Manhattan distance.
Optimized Additive with DA + mfcc
Additive resynthesis optimized with Dual Annealing using MFCC distance.
Optimized Additive with DA + pearson
Additive resynthesis optimized with Dual Annealing using Pearson correlation.
Optimized Additive with DA + spectral_convergence
Additive resynthesis optimized with Dual Annealing using spectral convergence.
Results below are from a batch run (maxiter=500) across metrics using Basin Hopping.
Optimized Additive with BH + cosine
Additive resynthesis optimized with Basin Hopping using cosine similarity.
Optimized Additive with BH + euclidean
Additive resynthesis optimized with Basin Hopping using Euclidean distance.
Optimized Additive with BH + itakura_saito
Additive resynthesis optimized with Basin Hopping using Itakura–Saito divergence.
Optimized Additive with BH + kl
Additive resynthesis optimized with Basin Hopping using KL divergence.
Optimized Additive with BH + manhattan
Additive resynthesis optimized with Basin Hopping using Manhattan distance.
Optimized Additive with BH + mfcc
Additive resynthesis optimized with Basin Hopping using MFCC distance.
Optimized Additive with BH + pearson
Additive resynthesis optimized with Basin Hopping using Pearson correlation.
Optimized Additive with BH + spectral_convergence
Additive resynthesis optimized with Basin Hopping using spectral convergence.
