Skip to content
This repository has been archived by the owner on May 17, 2019. It is now read-only.

Tooling and datasets for neural-network powered audio feature based synthesis

Notifications You must be signed in to change notification settings

hughrawlinson/feature-synthesis

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Feature Based Extraction Dataset Descriptions

Synth 1

A subsynth of the form [sawtooth, triangle, square] => [gain1, gain2, gain3] => lowpassfilter, with the parameters [gain1,gain2,gain3] and output ['rms'];

Datasets

  • s1: rms, each of the gains described above

Synth 2

A more fully featured subsynth. 6 oscillators, just a double of the above, exposing parameters for the filter (q and frequency), as well as detune params on 3 of the oscillators

Datasets

  • s2: Feature data includes [rms, spectral centroid, spectral rolloff], plus the normalised parameter values
  • s5: Same as above, but s/rolloff/flatness.

Synth 3

An FM synth, exposing parameters for the volume of each oscillator. Has a carrier, a modulator, and a higher order modulator.

Datasets

  • s3: Feature data includes [rms, spectral centroid, spectral rolloff] (may be broken).
  • s4: An attempt to fix s3. Also currently broken.