This script is inspired by Vincent Lostanlen's Synopsis Seriation paper. It filters stereo .wav file into a defined number of frequency bands, performs wavelet scattering analysis on each band, divides each band into wavesets of a defined number of zero-crossings, then rearranges those wavesets based on nearest neighbor matching on 200-waveset clusters before finally recombining the frequency bands together into an output .wav file. The result is a wide-stereo granular cloud that uses timbral elements from the input audio.

Further work is needed on the wavelet scattering and matching parts of the algorithm.