Each dataset has a unique set of syntheis parameters that are grid-sampled at the specified resolutions to produce the data files as described for each entry below. A representive set of audio samples from each datset can be auditioned from the 'Audio samples' drop down menu. The naming convention for example sounds is to concatenate the parameters with an =L, =M, and =H indicating that a low, medium, or high range value of the corresponding parameter was used to synthesize the sample.

Creating the the default dataset on your local system is a one-step process. You can also customize the dataset by editing the json configuration file (see the accompanying README file.)

** Description: ** * Amplitude modulation, (.5+.5*ππΌ*(πππ (2π ππ π‘)))*π ππ(2π ππ π‘) - modulating the amplitude of a sin wave with carrier frequency in [33, 660] Hz with a raised cos wave with a frequency in [0,55] Hz, ranging from slow shallow tremelo to the edge of the perception of a timbre. *

- cf_exp : 13 values mapped to n in [0, 1], where n -> 330*2^n, carrier frequency in Hz
- mf_exp : 21 values mapped to n in [0, 20], where n -> (1.22185)^n, modulator frequency in Hz (n=20 -> 55 Hz).
- mI : 11 values mapped to n in [0, 1], where n -> modulation index

1 audio variations per parameter setting for 3003 Parameters (3003 Files)

1 Hrs:40 Min:6.0 Sec of Audio segmented as 2.0 sec chunks

Total Size of Dataset: 183.29 MB @ SR 16Khz

** Description: ** * Frequency modulation of a sine wave with cf in [330, 660] Hz modulated by a sine wave in [0, 55] Hz. The parameters span the range from perceptual vibrato speeding up until it can just be heard as a timbre. sin(2pi*cf*t + mI*sin(2pi*mf*t)) *

- cf_exp : 13 values mapped to n in [0, 1], where n -> 330*2^n, carrier frequency in Hz
- mf_exp : 21 values mapped to n in [0, 20], where n -> (1.22185)^n, modulation frequency in Hz (n=20 -> 55 Hz).
- mI : 11 values mapped to n in [0, 25], where n -> modulation index

1 audio variations per parameter setting for 3003 Parameters (3003 Files)

1 Hrs:40 Min:6.0 Sec of Audio segmented as 2.0 sec chunks

Total Size of Dataset: 183.29 MB @ SR 16Khz

** Description: ** * A sequence of sine wave chirps rising or falling (nocts) with various rates and regularity in event timing. *

- irreg_exp : 3 values mapped to n in [0, 2], where n -> (n/event-per-second) as standard deviation of gaussian around regularly spaced events
- rate_exp : 5 values mapped to n in [1, 4], where n -> 2^(n) events per second
- nocts : 4 values mapped to n in [-3, 3], where n -> number of octaves spanned by chirp centered at center frequency
- evdur : 5 values mapped to n in [0.04, 0.2], where n -> duration of each chirp in seconds
- cf_exp : 7 values mapped to n in [0, 1], where n -> 440*2^(n), center frequency in Hz

1 audio variations per parameter setting for 2100 Parameters (2100 Files)

2 Hrs:20 Min:0.0 Sec of Audio segmented as 4.0 sec chunks

Total Size of Dataset: 256.35 MB @ SR 16Khz

** Description: ** * Short pops contructed of 3 noise samples follwed by narrow band pass filters are generated with a range of center frequencies, rates, and regularity. *

- irreg_exp : 11 values mapped to n in [0, 1], where n -> (n/event-per-second) as standard deviation of gaussian around regularly spaced events normalized by events-per-second.
- cf : 13 values mapped to n in [440, 880], where n -> Center frequency in Hz.
- rate_exp : 10 values mapped to n in [1, 4], where n -> 2**n events per second.

1 audio variations per parameter setting for 1430 Parameters (1430 Files)

1 Hrs:35 Min:20.0 Sec of Audio segmented as 4.0 sec chunks

Total Size of Dataset: 174.56 MB @ SR 16Khz

** Description: ** * Events times of varying temporal regularity are occupied by either a wood or a metal tok of varying ratios. The metal sound has a variable range of center frequencies. *

- hitratio : 6 values mapped to n in [0.5, 1], where n -> proportion of events times that will have a sound event
- wmratio : 9 values mapped to n in [0, 1], where n -> proportion of sound events that will be wood as opposed to metal
- rate_exp : 7 values mapped to n in [1, 3], where n -> 2^(n) events per second

1 audio variations per parameter setting for 378 Parameters (378 Files)

0 Hrs:37 Min:48.0 Sec of Audio segmented as 6.0 sec chunks

Total Size of Dataset: 69.21 MB @ SR 16Khz

** Description: ** * Regularly spaced 2-strike tapping. Similar to 'Tapping 1-2.wav' from McDermott and Simoncelli (2011 Neuron paper), but with systematic variation of rate and phase of 2nd tap. *

- rate_exp : 8 values mapped to n in [0.25, 2.25], where n -> 2**n events per second.
- phaserel : 10 values mapped to n in [0.05, 0.5], where n -> phase (in [0,1]) of the second tap in the cycle

1 audio variations per parameter setting for 80 Parameters (80 Files)

0 Hrs:13 Min:20.0 Sec of Audio segmented as 10.0 sec chunks

Total Size of Dataset: 24.41 MB @ SR 16Khz

** Description: ** * An 8-cycle 'engine.' The irreglular timing of a cycle of 8 pistons is repreated exactly, but the piston burst is radnom noise. Different samples have different irregular timing patterns for the same rate parameter. *

- rate_exp : 5 values mapped to n in [4, 8], where n -> 2^n events per second
- irreg_exp : 4 values mapped to n in [0.3, 0.9], where n -> normalized sd of gaussian around regularly spaced events: .1*n*np.power(10,n) = irregularity, gaussian sd=irregularity/eps

10 audio variations per parameter setting for 20 Parameters (200 Files)

0 Hrs:13 Min:20.0 Sec of Audio segmented as 4.0 sec chunks

Total Size of Dataset: 24.41 MB @ SR 16Khz

** Description: ** * Feedback comb-filtered noise: π¦[π]=(1βπΌ)π₯[π]+πΌπ¦[πβπΎ]. Intended for testing models that perform differently depending on how pitched a signal is. *

- pitchedness : 11 values mapped to n in [0, 1], where n -> 1-1/(2^4n) = alpha, the feedback coefficient. Thus n in [0,1] ->[whitenoise, pitched]

3 audio variations per parameter setting for 11 Parameters (33 Files)

0 Hrs:01 Min:22.5 Sec of Audio segmented as 2.5 sec chunks

Total Size of Dataset: 2.52 MB @ SR 16Khz

** Description: ** * Chua Oscillator varying alpha, beta, gamma of canonical dimensionless equations (see for example, Ch 3 of Bilotta, E., & Pantano, P. (2008). A Gallery Of Chua Attractors, World Scientific. - This dataset is derived from the DE64 example in that book). This set could be challenging to model because of the nonlinear relationship between parameters and audio. Although the parameters are grid sampled, the audio data characteristics are not balanced. The parameters explore the space between pitched and noisy sounds. A few sounds are not strictly textures because they are bistable and make a single switch from one from one to the other attractor. If you look at DSChua.py, you'll see a datastructure with 150 parameter sets for different Chua attractors. You can generate datasets for them by editing the config_file. *

- alpha : 9 values mapped to n in [6.2, 7], where n -> coefficient
- beta : 9 values mapped to n in [10.25, 11], where n -> coefficient
- gamma : 9 values mapped to n in [-0.2, -0.045], where n -> coefficient

1 audio variations per parameter setting for 729 Parameters (729 Files)

0 Hrs:36 Min:27.0 Sec of Audio segmented as 3.0 sec chunks

Total Size of Dataset: 66.74 MB @ SR 16Khz

** Description: ** * A mosquito or fly like sound that moves in space. Uses simplex noise for smooth modulations (eg distance) affecting frequency and amplitude at longish time scales. See the configuration to explore a wealth of synth parameters that go well beyond the mosquito-like sound. *

- cf_exp : 7 values mapped to n in [-1.5, 0], where n -> 440*2**n, the center frequency of the beating wings
- motion_freq : 3 values mapped to n in [1, 2], where n -> the lp filter cutoff frequency used to filter the noise that produces the smoothly-varying distance of the bug to you.
- busybodyFreq : 3 values mapped to n in [1, 15], where n -> synchronized noisy AM & FM average modulation frequency

1 audio variations per parameter setting for 63 Parameters (63 Files)

0 Hrs:08 Min:24.0 Sec of Audio segmented as 8.0 sec chunks

Total Size of Dataset: 15.38 MB @ SR 16Khz

** Description: ** * A wind sound generated with noise passed through a LP and BP filter modulated with simplex noise. *

- strength : 11 values mapped to n in [0, 1], where n -> stength (180_440*n is average freq of noise filter center frequency)
- gustiness : 11 values mapped to n in [0, 1], where n -> frequency parameter for simplex noise creating variation in filter center frequency and gain
- howliness : 11 values mapped to n in [0, 1], where n -> .5+40*n, the Q value for the BP noise filter.

1 audio variations per parameter setting for 1331 Parameters (1331 Files)

2 Hrs:35 Min:17.0 Sec of Audio segmented as 7.0 sec chunks

Total Size of Dataset: 284.33 MB @ SR 16Khz

** Description: ** * Wind chimes that can be sized and blown in wind of varying strength. Wind blows in the background, change this in the config file if you wish. *

- strength : 11 values mapped to n in [0, 1], where n -> stength (180_440*n is average freq of noise filter center frequency)
- chimeSize : 11 values mapped to n in [0, 1], where n -> size of the chime (inversely proportional to frequency)

1 audio variations per parameter setting for 121 Parameters (121 Files)

0 Hrs:14 Min:7.0 Sec of Audio segmented as 7.0 sec chunks

Total Size of Dataset: 25.85 MB @ SR 16Khz

** Description: ** * Like peepers (tree frogs), these sounds are made of sequences of chirps ('words'), with chirps varying in center frequency, range, and event period. The default dataset parameters are the number of peepers in the chorus, and the number of octaves swept by each chrip. *

- numPeepers_exp : 7 values mapped to n in [0, 6], where n -> 2^n number of Peepers to play simultaneously
- mean_nocts : 11 values mapped to n in [-1.25, 1.25], where n -> the mean number of octaves around a center frequency of each chirp in a word

1 audio variations per parameter setting for 77 Parameters (77 Files)

0 Hrs:08 Min:59.0 Sec of Audio segmented as 7.0 sec chunks

Total Size of Dataset: 16.45 MB @ SR 16Khz

** Description: ** * A swarm of bugs (although parameters for frequency excursions take sounds out of bounds of realsim). Uses simplex noise for smooth modulations (eg distance) affecting frequency and amplitude at longish time scales. See the configuration for other synth parameters to explore. *

- busybodyFreqFactor : 11 values mapped to n in [0, 0.5], where n -> noisy FM max variation (cf frequency variation in octaves)
- cf_exp : 11 values mapped to n in [-2, 0], where n -> 440*2**n, the center frequency of the beating wings

1 audio variations per parameter setting for 121 Parameters (121 Files)

0 Hrs:08 Min:4.0 Sec of Audio segmented as 4.0 sec chunks

Total Size of Dataset: 14.77 MB @ SR 16Khz

** Description: ** * Rough imitation of applause varying in number of clappers and clapping rate. Uses a bit of reverb that can also be manipulated in the config file. *

- numClappers_exp : 11 values mapped to n in [0, 7], where n -> round(2^n) number of clappers to play simultaneously
- rate_exp : 7 values mapped to n in [0, 3], where n -> 2^n claps (on average) per second per clapper

1 audio variations per parameter setting for 77 Parameters (77 Files)

0 Hrs:03 Min:51.0 Sec of Audio segmented as 3.0 sec chunks

Total Size of Dataset: 7.05 MB @ SR 16Khz