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) http://www.nongnu.org/fluid/
NumPy 1.0 or later (for some features) http://numpy.scipy.org/
Download the latest version from GitHub here: https://github.com/nwhitehead/pyfluidsynth/archive/master.zip
pyFluidSynth is packaged as Python source using distutils. To install, run the following command as root:
python setup.py install
For more information and options about using distutils, read: https://docs.python.org/2/distutils/
Here is a program that plays a chord for a second.
import time import fluidsynth fs = fluidsynth.Synth() fs.start() 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) time.sleep(1.0) fs.noteoff(0, 60) fs.noteoff(0, 67) fs.noteoff(0, 76) time.sleep(1.0) fs.delete()
First a Synth object is created to control playback.
start() method starts audio output in a separate thread.
To get sound, you need to choose an instrument. First load a
sfload(), then select a bank and preset with
program_select(track, soundfontid, banknum, presetnum)
To start a note, use the noteon() method.
noteon(track, midinum, velocity)
To stop a note, use noteoff().
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
generate the next chunk of audio, call
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
2 * len.
To join arrays together, use
To convert an array of samples into a string of bytes suitable for sending
to the soundcard, use
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 = pa.open( 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)) fl.delete() samps = fluidsynth.raw_audio_string(s) print len(samps) print 'Starting playback' strm.write(samps)
USING THE SEQUENCER
You can create a sequencer as follows:
import fluidsynth seq = fluidsynth.Sequencer()
This will by default create a sequencer that will advance at
a rate of 1000 ticks per second. To change the rate at which
the sequencer advances, you can give it the optional
parameter. As a clock source, it will use your system clock. In
order to manually advance the sequencer, you can give it the
use_system_timer=False. You will then have to advance
In order to make the sequencer aware of your synthesizer, you have to register it:
fs = fluidsynth.Synth() # init and start the synthesizer as described above… synthID = seq.register_fluidsynth(fs)
You have to keep the ID and use it as a
target for the midi events
you want to schedule. Now, you can sequence actual notes:
seq.note_on(time=500, absolute=False, channel=0, key=60, velocity=80, dest=synthID)
If you use relative timing like above, the sequencer will
schedule the event the specified time from the current position.
True (the default), you have to use
absolute track positions (in ticks). So the following code snippet
will do the same as the one above:
current_time = seq.get_tick() seq.note_on(current_time + 500, 0, 60, 80, dest=synthID)
You can also register your own callback functions to be called at certain ticks:
def seq_callback(time, event, seq, data): print('callback called!') callbackID = sequencer.register_client("myCallback", seq_callback) sequencer.timer(current_time + 2000, dest=callbackID)
Note that event and seq are low-level objects, not actual python objects.
You can find a complete example (inspired by this one from the fluidsynth library) in the test folder.
BUGS AND LIMITATIONS
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
<firstname.lastname@example.org> and Christian Romberg
Released under the LGPL v2.1 or any later version (this is the same as FluidSynth).
Copyright 2008--2015, Nathan Whitehead