Python bindings for FluidSynth
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 22 commits ahead, 22 commits behind nwhitehead:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Python bindings for FluidSynth

This module contains python bindings for FluidSynth. FluidSynth is a software synthesizer for generating music. It works like a MIDI synthesizer. You load patches, set parameters, then send NOTEON and NOTEOFF events to play notes. Instruments are defined in SoundFonts, generally files with the extension SF2. FluidSynth can either be used to play audio itself, or you can call a function that returns chunks of audio data and output the data to the soundcard yourself. FluidSynth works on all major platforms, so pyFluidSynth should also.


FluidSynth 1.0.7 (or later version) (earlier versions probably work, untested)

NumPy 1.0 or later (for some features)


Download the latest version from GitHub here:


pyFluidSynth is packaged as Python source using distutils. To install, run the following command as root:

python install

For more information and options about using distutils, read:


Here is a program that plays a chord for a second.

import time
import fluidsynth

fs = fluidsynth.Synth()

sfid = fs.sfload("example.sf2")
fs.program_select(0, sfid, 0, 0)

fs.noteon(0, 60, 30)
fs.noteon(0, 67, 30)
fs.noteon(0, 76, 30)


fs.noteoff(0, 60)
fs.noteoff(0, 67)
fs.noteoff(0, 76)



First a Synth object is created to control playback. The start() method starts audio output in a separate thread.

To get sound, you need to choose an instrument. First load a SoundFont with sfload(), then select a bank and preset with program_select().

program_select(track, soundfontid, banknum, presetnum)

To start a note, use the noteon() method.

noteon(track, midinum, velocity)

To stop a note, use noteoff().

noteoff(track, midinum)


You can also manage audio IO yourself and just use FluidSynth to calculate the samples for the music. You might do this, for example, in a game with WAV sound effects and algorithmically generated music. To do this, create the Synth object but don't call start(). To generate the next chunk of audio, call get_samples().


The length you pass will be the number of audio samples. Unless specified otherwise, FluidSynth assumes an output rate of 44100 Hz. The return value will be a Numpy array of samples. By default FluidSynth generates stereo sound, so the return array will be length 2 * len.

To join arrays together, use numpy.append().

To convert an array of samples into a string of bytes suitable for sending to the soundcard, use fluidsynth.raw_audio_string(samples).

Here is an example that generates a chord then plays the data using PyAudio.

import time
import numpy
import pyaudio
import fluidsynth

pa = pyaudio.PyAudio()
strm =
    format = pyaudio.paInt16,
    channels = 2, 
    rate = 44100, 
    output = True)

s = []

fl = fluidsynth.Synth()

# Initial silence is 1 second
s = numpy.append(s, fl.get_samples(44100 * 1))

sfid = fl.sfload("example.sf2")
fl.program_select(0, sfid, 0, 0)

fl.noteon(0, 60, 30)
fl.noteon(0, 67, 30)
fl.noteon(0, 76, 30)

# Chord is held for 2 seconds
s = numpy.append(s, fl.get_samples(44100 * 2))

fl.noteoff(0, 60)
fl.noteoff(0, 67)
fl.noteoff(0, 76)

# Decay of chord is held for 1 second
s = numpy.append(s, fl.get_samples(44100 * 1))


samps = fluidsynth.raw_audio_string(s)

print len(samps)
print 'Starting playback'


Not all functions in FluidSynth are bound.

Not much error checking, FluidSynth will segfault/crash if you call the functions incorrectly sometimes.


Original code by Nathan Whitehead <>. Contributions by Bart Spaans <> and Francisco J. Martinez <>.


Released under the LGPL v2.1 or any later version (this is the same as FluidSynth).

Copyright 2008--2015, Nathan Whitehead