Using statistic methods to synthesize sound of textures from white noise.
This is a simple implementation for the algorithm described in
Syhthesis of sound textures with tonal components using summary statistic and all-pole residual modeling
Hyung-Suk, Kim and Julius Smith
DAFx-16
This repository requires following packages:
- python 2.7
- numpy
- essentia
- librosa
- gammatone
usage: SoundTextureSynth.py [-h] [-i INPUT_PATH] [-o OUTPUT_NAME]
[-l OUTPUT_LENGTH] [-fs SAMPLE_RATE]
[-it ITER_TIME] [-lr LEARNING_RATE]
required arguments:
-i INPUT_PATH path to input file (source audio)
optional arguments:
-h
-o OUTPUT_NAME name of output file (default = 'out.wav')
-l OUTPUT_LENGTH length of output file(in seconds) (default = 5)
-fs SAMPLE_RATE sample rate (default = 44100)
-it ITER_TIME Maximum iteration time for gradient decent(in seconds) (default = 60)
-lr LEARNING_RATE learning rate for gradient decent (default = 0.3)
- Cross corelation faetures.
BSD