computer music with python
Python Other

Pippi: Computer music with python

v2.0.0 - Alpha 8

Installation and setup

Pippi requires python 3 which can be found here:

I'm just testing against the latest stable version (currently 3.6.2) at the moment.

I recommend you install pippi and work from a python virtualenv, which is built into the standard library in recent versions of python.

Create a virtualenv in a directory called venv with python 3.6+:

python3 -m venv venv

Activate the virtualenv:

source venv/bin/activate

Install from pip

To use the most recent release from pip, just:

pip install pippi

...from inside your virtualenv.

Install from source

Or to install the most recent development version, clone this repo and install with the -e / 'editable' option. (Also note the . pointing to the current working directory.)

pip install -e .

Pippi uses numpy, pysndfile, and Cython which all may create installation roadbumps depending on your platform.

If you're lucky, the above command ran without errors and you're good to go!

Otherwise, if there was a problem installing numpy please check out this page for more information about how to install it on your platform:

If there was a problem installing pysndfile you likely just need to install the library it wraps, libsndfile.

On linux, you should be able to find libsndfile in your distro's package manager.

If you're on a mac, it looks like it can be installed with homebrew:

On windows, Erik de Castro Lopo provides an installer on the libsndfile page:

If you have trouble compiling the Cython parts of pippi, you may need to install the python development headers, and/or build tools for your system. More here:

Please let me know if you run into problems!

Run the multi snare bounce example

cd examples

Which will produce a WAV file named multi_snare_bounce_example.wav in the examples directory.

Now what?

There are more examples, give em a whirl, and try your own. (And let me know when things break please!)

To run tests

make test

Release Notes

2.0.0 - Alpha 9 (in development)

Feature release, mostly:

  • Added 2d wavetable synthesis (similar to max/msp 2d.wave~) to Osc plus example script
  • Read wavetables from 1 channel sound files with wavetables.fromfile
  • Added a helper for async rendering with multiprocessing.Pool
  • SoundBuffers can now be pickled (enables passing them between processes)
  • SoundBuffer can be initialized (and spread across channels) from a normal python list
  • More interpolation modes (in progress)
  • Support pysndfile extra options for reading chunks / offsets from disk (todo)
  • Unit conversion helpers for time (todo)

2.0.0 - Alpha 6-8

This was meant to be a feature-only release, to add the final round of features before going into beta / bugfix mode. Instead I switched from using numpy arrays directly to a first pass of a more general typed memoryview approach, and moved some more things into cython.

Also, hoo boy was pip install pippi ever broken. It should be working now.

2.0.0 - Alpha 5

Bugfix release. Fixed an idiotic regression in SoundBuffer.

2.0.0 - Alpha 4

New in alpha 4:

  • More speed improvements!
  • Linear interpolation option for pitch shifting
  • Pulsar synthesis with Osc
  • Support for importing Scala .scl tuning files (Mapping file support coming later...)
  • Waveform visualization with graph.waveform
  • SoundBuffer.fill returns a copy of the sound instead of altering it in place
  • Sampler abstraction for Osc-like treatment of samples and banks of samples

2.0.0 - Alpha 3

Optimizations and improvements in this release:

  • Much better performance for wavetable and granular synthesis
  • Improvements and additions to the rhythm modules
    • Better handling of rhythm.curve which now takes a length param instead of an obscure combination of multipliers
    • rhythm.curve can now be provided a custom wavetable
    • Added MPC swing helper for onset lists (via rhythm.swing)
    • Added a euclidean rhythm generator
    • Added a pattern generation helper rhythm.pattern
    • Added pattern-to-onset and string-to-pattern helpers
      • Patterns are the same as pippi 1 (I may even just port some code) and can be in a few forms:
        • String literals with ascii notation eg: 'xx x- x' which is the same as 'xx.x-.x'
        • Lists of 'truthy' and 'falsey' values eg: ['1', True, 0, False] which is the same as 'xx..'
  • Misc bugfixes:
    • Fix random param for wavetable.window and wavetable.window
    • Fix bad params for wavetable.window and wavetable.window -- falls back to sine in both cases

2.0.0 - Alpha 2

This release includes a few missing pieces to core functionality including:

  • A crude squarewave wavetype for the wavetable osc!
  • Custom wavetables for the wavetable osc and window/wavetable generators! See the example for use with the wavetable osc.
  • A simple non-interpolating speed method on SoundBuffer for pitch shifting sounds
  • A set of rhythm helpers in the rhythm module useful for constructing onset / timing lists.
    • rhythm.curve which lets you map any of the window types to a list of onsets -- check out the example in the examples directory.
  • Some more example scripts including:
    • Demoing the rhythm.curve helper
    • A more interesting variation on the snare bounce example
    • Showing a user-defined wavetable used with Osc
    • Using the tune module with Osc to create a simple chord progression

2.0.0 - Alpha 1

This is the initial alpha release of pippi 2 -- which is very barebones at the moment, but already pretty functional!

Beware: the behavior of core functionality and features will probably change throughout the alpha releases of pippi 2. I'll try to document it here in the release notes.

This release provides:

  • SoundBuffer abstraction for reading/writing soundfiles and doing basic operations on sounds.
  • Osc abstraction for simple wavetable synthesis.
  • Initial set of built-in wavetables for windowing (sine, triangle, saw, inverse saw) and synthesis (sine, cosine, triangle, saw inverse saw)
  • Set of panning algorithms and other built-in sound operations like addition, subtraction, multiplication, mixing (and operater-overloaded mixing via sound &= sound), dubbing, concatenation.
  • A small set of helpers and shortcuts via the dsp module for loading, mixing, and concatenating (via dsp.join) sounds.
  • Basic granular synthesis and wavetable synthesis examples.

Some basic things from pippi 1 that are still missing (and coming very soon!):

  • Pitch shifting sounds
  • Square and impulse wavetables for windowing and synthesis
  • Using arbitrary wavetables for windowing and synthesis

Bigger things from pippi 1 that are still missing:

  • Pulsar synthesis
  • The interactive / live mode and console which includes stuff like
    • MIDI support
    • Voice handling
    • Live code reloading

I'm excited to revisit the interactive console especially, and I'm looking into some cool stuff like using pippi instruments as generators to feed into sound streams, which should hopefully overcome some awkwardness of treating pippi 1 instruments like streams.