A clojure music notation DSL for midi programming.
This Software is ALPHA QUALITY! Please file bug reports! Thanks!
Make sure you have boot-clj installed, e.g. brew install boot-clj. Then do:
git clone https://github.com/benwbooth/semitone.git
cd semitone
boot build
boot repl
Now you can start making music!
(play `(c d e f g a b c >))
semitone is a DSL language written in clojure specifically designed for writing MIDI-based music compositions. It is designed to allow users to compose music in a traditional music notation style, but being a DSL, it is flexible enough to allow any kind of algorithmic music generation as well. semitone makes use of the built-in Java real time sequencer code for MIDI sequencing and playback (See Big Caveat section at bottom of document). When users call the (play) function, the DSL language is converted into MIDIEvent objects, which are put into the Java sequencer's Track objects.
semitone takes inspiration from other music notation languages, such as MML, OpusModus, and Zel. Here is the structure of a "note" in semitone:
Length Note Parameters
When composing notes in semitone, the ordering of these three attributes is important. The length must come before the note, and the note must come before the parameters. Any of the three attributes may be omitted, however, in which case the attribute used in the previous note will remain in effect. This concept is similar to OpusModus Notation, except that the dynamic and expression attributes are combined into one parameters attribute.
The Length argument describes the duration of the note. Note length may be specified using Clojure's ratio literals (e.g. 1/4 is a quarter note), or using letters to designate various durations:
x: double-whole note
w: whole note
h: half note
q: quarter note
i: eighth note
s: sixteenth note
t: thirty-second note
j: sixty-fourth note
l: 128th note
m: 256th note
n: 512th note
o: 1024th note
Length letters can be combined together to form the sum of the duration of all length notes, e.g. wh = 1/1 + 1/2. If using ratio literals for the length, keep in mind clojure automatically tries to reduce ratio literals into integers when possible. I've included a ~(ratio numerator denominator) function which might help with this, but it's kinda verbose. A length prefixed with a - indicates a rest note. Dotted notes are supported. If you add a number to the end of a length, e.g. q3, the note length will be divided by 3. This is useful for simulating tuplets, although semitone doesn't adhere to all of the irregularities of traditional music notation regarding tuplets. It simply divides the note length by whatever number you gave.
There are two "repetition" symbols: = and -. The = symbol repeats the last note with the same duration. You can extend the duration by repeating the = symbol: === repeats the last note with the duration 3x longer. - is a rest repetition symbol. It rests for the duration of the previous note/rest. You can also extend the - rest symbol: --- plays a rest for 3x the previous note/rest's duration. The repetition operators may be useful for writing percussion.
A "note" in semitone can basically mean any MIDI message. Note On/Off pairs can be specified using the note name:
C
C#
db
f##
gbb
G
Notes written in uppercase will play one octave lower than lowercase notes. Accidentals may be given after the note name:
# sharp
## double-sharp
b flat
bb double-flat
n natural
Notes may also be specified using raw MIDI key values from 0-127:
k60
k72
Ties can be written by appending or prepending a _ character to a note.:
C#_ _C#
C#_ effectively plays a Note On without a Note Off, and _C# plays a Note Off without a Note On. Make sure your ties are properly paired, or you may get a note that plays forever!
If a note is prepended with a * character, it is treated as a Key Pressure message instead of a Note On/Off pair:
*c Send key pressure message which the note value C
*k60 Same as above
You may also append the octave number to the note:
c4 plays middle C.
"Notes" can also be CC messages:
cc32 plays a CC 32 message.
or Program changes:
prog1 plays a program change message that sets the program to 1.
or channel pressure:
*32 sets channel pressure to 32.
or pitch bend:
|-100 sets pitch bend to -100 cents from center position.
I tried to find a reasonable syntax for each of the MIDI Messages given in the specification.
Notes can be adjusted using "Parameter" syntax:
c4 !64 ;; plays middle c with attack set to 64 (range is 0-127)
c4 ?32 ;; plays middle c with release set to 32 (range is 0-127)
c4 !64?32 ;; plays middle c with attack set to 64 and release set to 32
attack refers to the MIDI Note On messages' value parameter, release is the Note Off's value parameter. There are two other parameters:
c4 $128 ;; plays middle c with the Note On's event time *displaced* forward by 128 ticks
c4 &-128 ;; plays middle c with the Note Off's event time *displaced* backward by 128 ticks
These are the displacement parameters. $ is displacement start. It adjusts the tick value of the Note On message. This could be useful for humanization. & is displacement end. It's useful for staccato/legato effects.
Parameter values can be given using integer values, as in the above examples, or using floating point literals. Parameters given with integer literals tend to have the raw MIDI values as their upper and lower bounds, e.g. 0-127. When parameters are given using floating point literals, the range tends to be from 0.0 to 1.0, with a few exceptions which will be noted.
All the parameter attributes I've given so far have been specified as absolute values, but you can also specify relative parameters using < and >:
c4 !64 >10 >10 >10 >10
The above will play middle c with an attack of 64, then will re-play the same note four more times, with the attack increasing by 10 every time. Here's the opposite, with the attack value decreasing:
c4 !64 <10 <10 <10 <10
Relative parameter adjustments may also be given by simply repeating the < and > symbols:
c4 !64 >>> >>> >>> >>>
The above plays middle c with attack 64, then plays four more times with the attack increasing by 3 every time.
One interesting thing about semitone is that just about every MIDI message is parameterizable. For example, you can set the value of CC32 to 64 like this:
cc32 64
But you can also do this:
cc32 64 70 74 80 84 90 94 100
or this:
cc32 0.0 0.1 0.2 0.3 0.4 0.5
You can also use the ~(env num-points values) function to create envelopes:
cc32 ~@(env 10 [0.0 0.5 -0.5 0.5])
env uses cubic interpolation to create 10 cc values using the values given as guides.
Almost all notes can be parameterized in this way:
- Note letters A-G expose the octave parameter
- key pressure messages expose the key pressure parameter
- cc messages expose the cc value parameter
- channel pressure messages expose the channel pressure parameter
- pitch bend messages expose the pitch bend parameter
- raw MIDI key messages (e.g. k60) expose the midi key parameter
- explicit attack parameters (e.g. !64) expose the attack parameter
- explicit release parameters (e.g. ?64) expose the release parameter
- explicit displacement-start parameters (e.g. $100) expose the displacement-start parameter
- explicit displacement-end parameters (e.g. &100) expose the displacement-end parameter
The idea being that, once a parameter is exposed, you can then give integer or decimal literals to tweak the parameters until the next note symbol is encountered. You can then also use the env function to easily write expressive envelopes.
Now that we (hopefully!) have a basic idea of how to write individual notes, how do we sequence them together to compose music?
In semitone notation, any notes surrounded by parentheses () play sequentially:
(c d e f g)
The parentheses notation can also be used to simulate measure bars:
(c d e f) (g a b c >)
Any notes surrounded by square brackets [] play simultaneously:
[c e g]
You can use the parentheses to create a sequence of notes, and you can use the square brackets to write both chords and separate voices.
The duration of a semitone expression in parentheses is the sum of the durations of its contents. The duration of an expression in square brackets is equal to the duration of the last element inside the square brackets.
In semitone notation, clojure keyword literals (e.g. :key-signature) followed by a value can be used to adjust all the stateful parameters that are used to translate notation into MIDI events:
:sequencer sequencer ;; set the sequencer object we're writing to
:displacement-start value ;; set the current default displacement-start value in ticks
:displacement-end value ;; set the current default displacement-end value in ticks
:attack value ;; set the current default attack value (0-127)
:release value ;; set the current default release value (0-127)
:channel-pressure pressure ;; set the current default channel pressure value (0-127)
:pitch-bend value ;; set the current default pitch bend value (-8192 to 8191)
:key-pressure value ;; set the current default key pressure amount (0-127)
:position pos ;; set the current tick position we're writing to (0-...)
:tie ;; state for keeping track of note ties
:key number ;; set the current default MIDI key value (0-127)
:message-type type ;; set the current default MIDI message type (ShortMessage/NOTE_ON, etc.)
:ch channel ;; set the MIDI channel we're writing to (0-15)
:track track ;; set the MIDI track we're writing to (0-...)
:key-signature {} ;; define a map of note name to semitone offsets to describe a key signature
:time-signature [4 4] ;; define a two-element vector describing the time signature (currently unused)
:transpose semitone ;; transpose the MIDI note messages
:octave number ;; set the current default octave number to use when no octave is given
:note-octave ;; state for keeping track of note octaves
:param param-keyword ;; set the current default tweakable parameter keyword
:cc-value value ;; set the current default CC value
:tempo bpm ;; set the current tempo in BPM
:length ;; set the current default note length in ticks
:repeat-length ;; set teh current default repeat length in ticks
:midi-message ;; internal state
:notes ;; sets the semitone expression being evaluated
You shouldn't need to use most of these keywords, but they can be useful for certain things, like setting the key signature and the tempo.
There are a few special variables defined in the semitone namespace:
*seq*is the default MIDI sequence object*synthis the default MIDI synthesizer object*sequencer*is the default MIDI sequencer object
There are also a few functions that are useful:
-
(ratio value)converts the value to a ratio object -
(env num-points values)produces an envelope with num-points points using values as a guide -
(load-soundfont soundfont synth)loads a synth with the given soundfont file -
(make-sequencer sequence synth)creates a new sequencer object -
(compose notes sequencer)writes thesemitonenotation to the sequencer -
(play notes sequencer)writes thesemitonenotation to the sequencer and plays the sequencer -
(clear sequencer)clears all the MIDI events from the sequencer
I also wrote a print-method for the *seq* object so you can see the raw MIDI events in a MIDI sequence for debugging purposes.
Let's now look back at the first simple example from the beginning of this document:
(play `(c d e f g a b c >))
We're calling the (play) function, which composes the semitone notation into a series of MIDI events, and then plays them through the default synthesizer. Notice that the semitone notation is always written using syntax quoting. I highly encourage you to write semitone notation using syntax quotes, because then you can use unquoting to insert the output of functions directly into your notation. This is how the env enveloping function works. It uses the syntax unquote operator ~@ to put the envelope parameters into the notation.
The > after the last c note increases the octave of the previous note using the octave parameter exposed by the c note and doing a relative increment of +1. This looks suspiciously similar to the > operator in MML language, but remember in semitone notation, the > operator adjust the previous note! Also, it is much more flexible because the > operator in semitone can adjust any parameter value, not just octaves! Everything can be tweaked.
Since this whole thing is implemented on top of Java's Sequencer APIs, it suffers from a limitation: external MIDI sync is completely unimplemented. That means you can't use this with any external sequencer (such as reaper) at the moment. I'm hoping to remedy this by patching the Java sequencer API to add external MIDI clock and MTC master/slave support. I've gotten started on it, but I don't know when it'll be finished. When it is though, you should be able to sync semitone with external sequencers and send and receive MIDI messages between them.
I'm a programmer not a composer, and I haven't had time to produce more tests/examples. Sorry about that! Patches welcome! :)
I know there are still LOTS of bugs in the code. Also, the code quality is pretty bad. I used regexes for parsing to keep things simple, but I'd like to convert it to use parser combinators at some point.
- Finish writing Java sequencer that can do external sync
- MPE support
- implement time signatures
- add shortcuts for intervals, chords, key signatures, dynamics
- convenient meta message syntax
- convenient sysex syntax
- SMPTE time?