Minimal algorithmic music composer and synthesizer
Go C Other
Switch branches/tags
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
audio audio: avoid null midi devices list, use empty list instead Nov 18, 2017
cmd/glitch glitch: update deps Jan 6, 2018
dist dist: add script to produce web variant Nov 18, 2017
examples examples: fix syntax changes Oct 8, 2017
ui glitch: update deps Jan 6, 2018
vendor/ glitch: update deps Jan 6, 2018
.gitignore updated gitignore files Jul 2, 2017
.travis.yml doc: change example links to the actual webview Oct 6, 2017
Gopkg.lock glitch: update deps Jan 6, 2018
Gopkg.toml vendor: init Oct 8, 2017
LICENSE add GPL license Sep 29, 2017 fixed pluck() documentation Apr 30, 2017
appveyor.yml ci: deploy appveyor on tags Nov 18, 2017


Build Status

Glitch is a minimal environment for creating algorithmic music and live coding.

It uses arithmetic expressions to synthesize instruments and create music patterns.

Try it online:

Download for Mac, Windows or Linux:

Read more about Glitch on Medium:


On linux: make alsa=1 or make alsa=1 pulse=1. If you want to use JACK: make jack=1.

On windows: make windows=1.

On MacOS: make macos=1.

Asm.js: make js (requires Docker).


Glitch syntax is arithmetic expressions, most likely you still remember it from the math class.

Arithmetics: + - * / % (modulo) ** (power)

Bitwise: & | ^ (xor or bitwise not) << >>

Compare: == != < <= > >= (return 1 or 0)

Grouping: ( ) , (separates expressions or function arguments)

Conditional: && || (short-circuit operators)

Assignment: = (left side must be a variable)


Function Description Example
sin(freq) sine wave at given frequency sin(440)
tri(freq) triangular wave at given frequency tri(440)
saw(freq) saw-tooth wave at given frequency saw(440)
sqr(freq, [pwm=0.5]) square wave at given frequency and (optionally) pwm sqr(440)
fm(freq, [m1, v1, m2, v2, m3, v3]) FM-synthesizer with 3 operators, vN is operator strength, mN is operator multiplier, operators 1 and 2 are parallel, operator 3 is sequential to operator 1 fm(440, 0.5, 0.5)
tr808(drum, [vol=1], [shift=0]) plays TR808 drum sample at given volume and pitch shift. The following drum IDs may be used: BD (bass drum), SD (snare drum), MT (middle tom), MA (maracas), RS (rimshot), CP (clap), CB (cowbell), OH (open hat), HH (hi-hat) tr808(BD, 1)
piano(freq) very basic piano sample at the given frequency piano(440)
pluck(freq, decay) Karplus-Strong string synthesizer, fill is a function used to prepare the initial values in the delay buffer pluck(440, 0.7)

FM synthesizer, TR808 sampler and Piano are reset if any of the parameters is NAN. All instruments return NAN if the input is NAN.

All instruments return a sound wave with the given frequency in the range [-1..1], so you can combine them by adding the signals (e.g. (sin(440)+sin(220))/2) or modulate using multiplication, e.g. saw(440)*sin(1).

You may put custom samples into the samples subdirectory, each group of samples should be in a separate folder. Then you could use samples providing the directory name as a function. For example if you have samples/bass/bass0.wav and samples/bass/bass1.wav you may call them as bass(0) and bass(1) respectively. Samples are expected to be in the WAV mono 16-bit format with 44100 Hz sample rate.


Function Description Example
a(i, ...) array element by its index, most primitive sequencer sin(a(t>>10, 440, 466, 493))
seq(tempo, ...) switches elements at given tempo sin(loop(120, 400, 466, 493))
loop(tempo, ...) switches elements at given tempo, unlike seq() it evaluates each argument for each time frame which makes it possible to nest loops sin(loop(60, seq(240,400,466,493), seq(480, 400,493)))

Tempo can be a single number (beats per minute) or a pair (offset, bpm), where offset is number of beats to skip before starting the sequence.

Seq and loop values can be pairs, too. Then the first value is a relative beat duration and the second is the actual returned value: seq(120, (3/4, 1), (1/4, 1)) returns the value of "1", but the first value lasts 3 times longer than the second value.

If seq takes more values, they will be sliding from one another, e.g. seq(120, (1, 0, 4, 2), (1, 2, 4, 0)) slides the values like 0->4->2->2->4->0. The first value in a group is still a relative beat duration.

Seq and loop return NAN every then the value is changed.


Function Description Example
r(max) random number in the range [0..max), it sounds like white noise, good for synthesizing drums or making randomized music patterns r(100)
s(phase) sine wave amplitude at the given phase, unline sin() you must provide phase in the range [0..1] s(t*14)
l(x) binary logarithm, useful to convert frequencies to note values note=l(440)*12
hz(note) note frequency of the given note index, index 0 is note A of 4th octave, you may also use helper variables like A#4, C2, Db3 sin(hz(A4))
scale(pos, mode) return note index at given position in given scale, scale 0 is major scale, scale 6 is minor sin(hz(scale(t>>11&7)))
env(signal, (dt, level)...) Creates an ADSR envelope for the signal, envelope is reset if signal is NAN, if first part has non-zero level - the initial level starts from 1, otherwise from 0; if last argument is not zero - the release section is inserted automatically env(v, (0.1, 0.2))
mix(...) mixes voices together, each parameter is a signal or a pair of (volume, signal). Signals are clipped if overflow occurs mix(sin(220), sin(440), tri(880))
lpf(voice, cutoff) applies low-pass filter to the voice at given cutoff frequency lpf(v, 200)
hpf(voice, cutoff) applies high-pass filter to the voice at given cutoff frequency hpf(v, 400)
bpf(voice, cutoff) applies band-pass filter to the voice at given cutoff frequency bpf(v, 400)
bsf(voice, cutoff) applies band-stop filter to the voice at given cutoff frequency bsf(v, 400)
delay(voice, time, level, feedback) delays signal by given time, delay level can be controlled as well as the amount of delay feedback, which affect the number of delay repetitions delay(v, 0.1, 0.5, 0.2)


To reuse the same expression multiple times you may create a macro:

$(organ, (sin($1)+0.4*sin($1+7)+0.3*sin($1-5))/3)

Macros are definde using the $(name, body) function. Body can consist of multiple expressions if you extra parenthesis, e.g. $(filter, (z=saw($1), lpf(z))).

There are special argument variables $1..$9 that get expanded to the actual values when the macro is called.

Special variables:

t is time variable that increases at rate 8000/second.

x and y in the web version are mouse cursor position, normalized to (0..1) range.

bpm is a tempo, user input is synchronized with the playback at this rate, so you might want to use bpm=120/4 to synchronize user input every 4 beats.


To apply the same expression to a number of variables you may use each() function. It takes a list of formal variables, a function, and a list of actual values: each(f, sin(f), 440, 880, 220).

You may pass multiple variables as well: each((vol, freq), vol*sin(freq), (1, 440), (0.4, 880), (0.2, 220)).

This is useful for live MIDI input.


Glitch provides special variables that change their values if a MIDI keyboard is used:

  • k0, k1, ..., k9 - MIDI key values (0=A4, 1=A#4 etc)
  • v0, v1, ..., v9 - MIDI velocity (it gradually fades out on key release)
  • g0, g1, ..., g9 - MIDI gate signal (it is set to NAN immediately on key release)

Here's how you can play a sine wave with a MIDI keyboard (handling up to 5 keys pressed at a time, use k5, k6 etc to get more polyphony):

each((k, v), v*sin(hz(k)),
  (k0, v0), (k1, v1), (k2, v2), (k3, v3), (k4, v4))

Special variables x and y are set to the pitch wheel and modulation wheel values if a MIDI keyboard is used.