splitRand

Simple python script to split audio at silences and reconcatenate into a series of new, randomly structured files

Introduction

splitRand.py is a python3 script that leverages the power of pydub and librosa to chop up a sound file based on silences detected within it. It then reorders the slices into new, random structures, as many times as you like. Nice for drum patterns, good for creating new sound structures for game audio libraries, great for making gobbledygook out of spoken word.

All segments generated have a short fade in and fade out applied to guarantee there are no sample-crossing clicks.

Installation and dependencies

To install, first get python3 installed on your machine.

Windows python3 instructions

download: https://www.python.org/downloads/windows/
installation reading: https://docs.python.org/3/using/windows.html

OSX python3 installation

install brew instructions here
then install python brew install python
further reading: https://docs.python-guide.org/starting/install3/osx/

install dependencies with pip3

pydub

pydub is an awesome toolkit developed by James Robert and explained here: http://pydub.com/

pip3 install pydub

librosa

I'm not sure if Librosa is really needed for this script, but as you might ultimately want to produce graphs of audio data or to make more sophsiticated splitting of audio, this might be a useful tool to have access to. Librsoa is explained here https://librosa.org/doc/latest/index.html.

pip3 install librosa

tqdm

This is not important but it helps you to see how much longer you've got before the process is completed. Can be reassuring for very long processes.

More information is available here: https://github.com/tqdm/tqdm

pip3 install tqdm

ffmpeg

Insanely useful media file manager, you'll need it anyway

brew install ffmpeg

think about where you'd like to keep this script and make a note of the path

once installed, note the path to this script so that you can call it on the commandline:

python3 path/to/splitRand.py --file=path/to/waveFile.wav

Download the package

Either download the zip of this repository or clone it:

git clone https://github.com/tinpark/spliRand.git

Arguments

Add arguments to refine the splitting process and define the number of new random files:

python3 path/to/splitRand.py --file=path/to/waveFile.wav --thresh=-60 --msl=10 --fdIn=5 --fdOut=10 --iters=16

The above will split your audio at whenever it's dropped to -60dB for 10ms or more, a fadein of 5ms and fade out of 10ms will be applied to all slices and 16 random iterations of the short segments files will be concatenated together to make 16 new .wav files.

example python calls

Try the following examples to test things.

First cd into the folder you downloaded or cloned from github.

Then run this line of code

python3 splitRand.py --file="examples/Herve_klezmer1.wav" --thresh=-60 --msl=5 --iters=4

python3 splitRand.py --iters=6 --file="examples/Drum160Fight.wav" --thresh=-10 --msl=5

python3 splitRand.py --iters-12 --file="examples/martin.wav" --thresh=-30 --msl=5

Credits:

Parts of this script were hacked from the following pages, many thanks especially to: