19.06.2018       Выпуск 235 (18.06.2018 - 24.06.2018)       Интересные проекты, инструменты, библиотеки

decadence - пишем музыку прямо в Vim


Экспериментальная функция:

Ниже вы видите текст статьи по ссылке. По нему можно быстро понять ссылка достойна прочтения или нет

Просим обратить внимание, что текст по ссылке и здесь может не совпадать.


Plaintext music tracker and midi shell w/ Vim integration >:) Open-source under MIT License (see LICENSE file for information)


Copyright (c) 2018 Grady O'Connell


Compose music in a plaintext format or type music directly in the shell. The format is vertical and column-based, similar to early music trackers, but with some modern features to make it more accessible and readable by musicians. The format is intended to be dense but readable once learned.

In a traditional tracker, individual notes would take place over multiple channels. For instance, a C major chord would be specified in a way a computer would understand it: 3 notes across 3 separate channels: C,E,G. In Decadence, you can do it as a chord in 1 channel: You could choose to write it as "Cmaj". or since we're in the key of C, "1maj", "maj", "major", "M", or "I" would also work.

Chord voicings can get complex and specific, and decadence's format prefers density, So you may find yourself writing voicings spanning octaves like this:


The above chord voicing spans 3 octaves and contains 9 notes. It is a Cmaj7 chord w/ an added #4, followed by a lower octave Csus2. Then at the bottom, there is a C bass note. Writing this in a tracker with 1 note per channel would be difficult to read (at least to me). As cryptic as it may seem to non-musicians, condensed chord voicings are going to make more sense to your ear over time than seeing random note letters fly by. For this reason, decadence prefers numbered note notation and key note transposition over arbitrary note names (even though both are valid). If no number prefix is given, like in this example, it is always 1, and unless we tranpose, that means C.

Note to musicians: There are a few quirks with the parser that make the chord interpretation different than what musicians would expect. For example, slash chords do not imply inversions, but are for stacking across octaves. Additionally, note names do no imply chords. For example, C/E means play a C note with an E in a lower octave, whereas a musician might interpret this as a specific chord voicing. (Inversions use jazz letter suffix or shift operator (majb or maj>))


Plaintext = Vim = Zen.

Music is code.


Decadence is a new project, but you can already do lots of cool things:

  • Strumming
  • Arpeggiation
  • Tuplets and polyrhythms
  • CC automation
  • Vibrato, pitch, and mod wheels
  • Dynamics
  • Accents
  • Velocity
  • Inversions
  • Midi channel stacking
  • Note length
  • Delays
  • Markers, repeats
  • Scales and modes by name


You can use this with General Midi out-of-the-box on windows but who wants to write music like that? We need VSTs!

For windows, you'll need a virtual midi driver, such as loopMIDI and a VST host or DAW.

If you're on Linux, you can use soundfonts through qsynth or use a software instrument like helm or dexed. VSTs should work here as well.

If you feed the MIDI into a DAW you'll be able to record the output through the DAW itself. I'm currently looking into recording via a headless host.

Command line parameters (use -):

- (default) starts midi shell
- (filename): plays file
- c: play a given sequence
    - Passing "1 2 3 4 5" would play those note one after another
- l: play a single line from the file
    - Not too useful yet, since it doesn't parse context
- +: play range, comma-separated (+start,end)
    - Line numbers and marker names work
- t: tempo
- x: grid
- n: note value
- c: columns
    - specify width and optional shift, instead of using auto-detect
    - positive shift values create a "gutter" to the left
    - negative values eat into the size of the first column
- p: set midi patches
    - command-separated list of patches across tracks
    - GM instruments names fuzzy match (Example: Piano,Organ,Flute)
- --sharps: Prefer sharps
- --solfege: Use solfege in output (input not yet supported)
- --flats: Prefer flats (currently default)
- --device=DEVICE: Set midi-device (partial match supported)

Global commands:

- %: set vars
- ;: comment
- :: set marker (requires name)
- @: go back to last marker, or start
- @@: pop mark, go back to last area
- @start: return to start
- @end: end song
- R: set scale (relative)
    - Scale and mode names suppored
- S: set scale (parallel)
    - Scale and mode names suppored
- P: set patch(s) across channels (comma-separated)
    - Matches GM midi names
    - Supports midi patch numbers
    - General MIDI name matching

Track commands

- ': play in octave above
    - repeat for each additional octave (''')
    - for octave shift to persist, use a number instead of repeats ('3)
- ,: play in octave below
    - number provided for octave count, default 1 (,,,)
    - for octave shift to persist, use a number instead of repeats (,3)
- >: inversion (repeatable)
- <: lower inversion (repeatable)
- ch: assign track to a midi channel
    - midi channels exceeding max value will be multiplexed to different outputs
- pc: program assign
    - Set program to a given number
- Global var (%) p is usually prefered for string matching
- cc: control change (midi CC param)
    - setting CC5 to 25 would be c5:25
- bs: bank select (not impl)
- ~: vibrato and pitch wheel
- `: mod wheel
- ": repeat last cell (ignoring dots, blanks, mutes, modified repeats don't repeat)
- *: set note length
    - defaults to one beat when used (default is hold until mute)
    - repeating symbol doubles note length
    - add a number for multiply percentage (*50)
- .: half note length
    - halfs note value with each dot
    - add extra dot for using w/o note event (i.e. during arpeggiator), since lone dots dont mean anything
    - add a number to do multiplies (i.e. C.2)
- !: accent a note (or set velocity)
    - set velocity by provided percentage
    - !! for louder notes
    - !! for louder accent
    - !! w/ number set future velocity
- ?: play note quietly (or set velocity)
    - repeat or pass value for quieter notes
- T: tuplet: triplets by default, provide ratio A:B for subdivisions
- ): delay: set note delay
- \: bend: (not yet implemented)
- &: arpeggio: plays the given chord in a sequence
    - infinite sequence unless number given
    - more params coming soon
- $: strum
    - plays the chord in a sequence, held by default
    - notes automatically fit into 1 grid beat

Note: Percentage values specified are formated like numbers after a decimal point:
Example: 3, 30, and 300 all mean 30% (read like .3, .30, etc.)

The Basics

If you're familiar with trackers, you may pick this up quite easily. Music flows vertically, with separate columns that are separated by whitespace or setting separators.

Each column is represents a track and they default to separate midi channel numbers. Tracks sequence notes. You'll usually play at least 1 track per instrument. This doesn't mean you're limited to just one note per track though, you can keep notes held down and play chords as you wish.

Each cell row in a track can contain both note data and associated effects.

By default, any note event in a track will mute previous notes on that track

The following will play the C major scale using numbered notation:

; Major Scale -- this is a comment, write whatever you want here!

; 120bpm subdivided into 2 (i.e., eighth notes)

%t120 x2


The tempo is in BPM, and the grid is based in subdivisions. Musicians can think of grid as fractions of quarter note, The grid is the beat/quarter-note subdivision.

Both Tempo and Grid can be decimal numbers as well.

Transposition and Octaves

Notice the bottom line has an extra apostrophe character ('). This plays the note in the next octave For an octave below, use a comma (,). You can use a number value instead to make the octave changes persistent (,2).

Holding Muting

Notes will continue playing automatically, until they're muted or another note is played in the same track.

You can mute all notes in a track with -

To control releasing of notes, use dash (-). The period (.) is simply a placeholder, so notes continue to be played through them.

; hold note 1 until next note (dots aren't notes, just empty placeholders for example)

; auto-mute by specifying note value (*):
; manually mute with '-'

Note durations can be controlled by adding * to increase value by powers of two, You can also add a fractional component to multiply this. The opposite of this is the dot (.) which halves note values

; set note based on percentage (this means 30%)

; set note based on percentage (33%)

; set note based on percentage (33.3%)

; etc...

Notes that are played in the same track as other notes mute previous notes. In order to overide this, hold a note by suffixing it with underscore (_).

A (-) character will then mute them all.

; Let's hold some notes


You can play notes individually or use chord names.

Let's play a scale with some chords:

%t120 g2

There are lots of chords and voicing (check the .py file under CHORDS and CHORDS_ALT) and I'll be adding a lot more. All scales and modes are usable as chords, so arpeggiation and strumming is usable with those as well.

Arpeggios and Strumming

Chords can be walked if they are suffixed by '&' Be sure to rest in your song long enough to hear it cycle.


After the &, you can put a number to denote the number of cycles. By default, it cycles infinitely until muted

The dollar sign is similar, but walks an entire chord or scale within a single grid space:


Velocity and Gain/Volume

Control velocity and volume of notes using the %v## or !## flags respectfully Example: %v0 in min, %v9 is 90%. But also: %v00 is minimum, %v99 is 99% (%v by itself is full)

Interpolation not yet impl



Tilda(~) does vibrato. Vibrato uses the mod wheel right now, but will eventually use pitch wheel oscillation.


Columns are separate tracks, line them up for more than one instrument

1,2  1
.    4
.    5
.    1'
.    4'
.    5'
.    1'2


still working on this feature, almost ready

':' sets marker and '@' loops to it.


Repeat counting, callstack, etc. coming shortly. Code almost done.


Very early support for this. See tuplet.dc example. The 't' command spreads a set of notes across a tuplet grid, starting at the first occurence of t in that group. Ratios provided will control expansion. Default is 3:4. If no denominator is given, it will default to the next power of two (so 3:4, 5:8, 7:8, 11:16). So in other words if you need a 5:6, you'll need to write t5:6. :) The ratio of the beat saves. You only need to specify it once per group. For nested tripets, group those by adding an extra 't'.

Consider the 2 tracks:

1     1t
2     2t
3     3t
1     1t
2     2t
3     3t

The spacing is not even between the sets, but the 't' value stretches them to make them even in a default ratio of 3:4

What's the plan?

Not everything is listed here because I just started this project. More to come soon!

Things I'm planning on adding soon:

- Improved chord interpretation
- MIDI input/output
- MIDI stabilization
- Headless VST rack integration
- Csound, Sonic Pi, libgme (chiptune lib) integration
- Text-to-speech and singing (Espeak/Festival)

Features I'm adding eventually:

- Recording and encoding audio output of a project
- Midi controller input and recording
- Midi input chord analysis

I may eventually rewrite this in C++ to achieve better speed. I'll make use of python's multiprocessing and possibly separate processes to achieve as much as I can do for timing critical stuff.


Лучшая Python рассылка

Разместим вашу рекламу

Пиши: mail@pythondigest.ru

Нашли опечатку?

Выделите фрагмент и отправьте нажатием Ctrl+Enter.

Система Orphus