DR. T Algorithmic Composer
posted 1 february 2002
Keyboard / October 1986 by Ted Greenwald
"I shall never believe that God plays dice with the world." -Albert Einstein
"I want to play a little dice with my music."-Emile Tobenfeld, a.k.a. Dr. T
-
JUST WHEN YOU THOUGHT IT WAS SAFE to ditch your Commodore 64 for one of two sexy new megamonster computers, Dr. T's puts out another
program. Not just another program, but one that does things no other program even comes close to. And not just things that aren't
available elsewhere, but things that many musicians will find take a fair amount of the drudgery out of creating contemporary
electronic music, and add a big dollop of fun. In combination with the updated Keyboard Controlled Sequencer (KCS) for the Commodore
128 - also covered in this report - the Algorithmic Composer delivers a kind of MIDI processing power only hinted at until now.
Both packages are extensive enough that we can't explain every detail here, and both were under revision throughout the review period,
so you can expect even more features than you read about below. We'll try to hit the high points.
-
Algoritmic Composer
The Algorithmic Composer package, consisting of three separate programs, applies logical operations to your input in order to
generate something approaching music. Doubtless some musicians will bristle at the thought of calling upon a computer to write their
music, while others will view this software as just what consciously or unconsciously they've been looking for: instant musicianship.
Well, check your prejudices at the disk drive door, friends; the world isn't quite so cut-and-dried.
The computer, in this case, just crunches input data while following some basic preprogrammed rules and adding some degree of
randomness something we all do, at so me expense of creative energy and spontaneity, while we make music. The rules are set, as they
are in any musical instrument from a violin to the most sophisticated turnkey music workstation, the degree of randomness is under
some control (which is more than many of us can manage), and ultimately it's up to the user to feed the instrument good information
if he or she wishes to avoid mediocre results. That takes Creativity, esthetic discrimination, a knowledge of musical materials, and
familiarity with the tools at hand-even in the computer age.
-
The Phrase program makes up melodies according to parameters you provide. The program uses something like a chord chart actually a
scale chart to decide what pitch-classes it can choose from during any given beat. Scale options are quite comprehensive: major;
harmonic, melodic, and natural minor; dorian, lydian, mixolydian, and phrygian modes; pentatonic; whole-tone; and four user defined
scales. (The latter can be entered in any order and can contain redundancies, giving you more sophisticated kinds of control over
certain operations than the program generally offers, should you choose to delve that deep). All scales cover three octaves,so that's
the outside range for melodies and melodic intervals. Melody range options are one, two, or three octaves, and melodic interval range
is adjustable in semitones.
The program breaks melodies up into a theme and a phrase. A theme, basically, is a motif which serves as the basis for the phrase.
You can set a theme length in beats/fractions-of-a-beat. The program then cycles through the theme, transposing it according to the
scale chart, in order to compute the phrase. (Phrase length is defined by the length of your initial scale chart) Theme and phrase
lengths tend to relate in the following ways: If the theme is the same length as the phrase, you'll get a meandering, through composed
melody without much coherence. If the theme issome integral division of the phrase (say, a half, a third, or a quarter as long),
you'll get something phrased pretty much like pop music, where a melodic fragment simply transposes with the harmonic changes.
If the theme length is some uneven division of the phrase, the melody will be phrased in a coherent, but less obvious, manner.
If at any point the scale chart takes the theme out of the defined melodic or intervallic range, the relevant thematic material will
be transposed into place on a note-by-note basis, which can lead to some interesting variations within the phrase.
In addition, you can specify an average number of notes per measure, a smallest note value (of which all other note-values will be
multiples), and a rest probability (percentage chance that any given quantization value, or multiple there of, will be filled with
silence), and you can choose among sets of rules for the handling of consonant scale steps and velocity values. If you come across
a melody you like and wish to introduce variations in a controlled fashion (rather than generating a new theme and phrase), there's
a scramble function that swaps your choice of pitches, note-values, or durations within a theme.
When you boot up Phrase, the default values give you instant perky diatonic techno-pop lines over standard blues changes. By messing
around just a bit, you can make it grind out techno-pop bass lines, lead lines, melodies, or backing patterns. lake it from us: It's
a kick. We found that the best way to gain control over the program, however, is to enter values that will cause it to emulate lines
from some specific composition that you're familiar with. This gets you directly into fine-tuning the parameters to get what you're
looking for, rather than allowing the program to pump out meaningless blatlher until it stumbles onto something interesting. When you
know what you're after, it's surprising how close you can come.
Much of the program's output is (predictably) lacking in character, and occasionally it spits a few notes fresh out of left field.
But it's all good, solid, editable MIDI data,so you can use a sequencer to take what you like and leave the rest (Algorithmic
Composer files can be read by Dr. T's Commodore 64 sequencer) or just hit the button again and generate some different material.
-
This program is pretty simple, but difficult to evaluate. It takes five lists of user-input values-one each for note-value, duration,
MIDI channel, pitch, and velocity and combines them into a single sequence. Here's a simple example to show how it works: Say you
enter three values for pitch, Middle C, D, and E, and two values for MIDI channel, 1 and 2. The result will be a sequence six events
long (three notes times two channek)-C played on channel 1, D on channel 2, E on 1, C on 2, D on 1, and E on 2. The series loop
independently, so after that, the sequence will repeat ad infinitum.
The program holds two sets of series data at a time. Thus it can generate two independent, simultaneous sequences, which
adds
contrapuntal possibilities to an already complex and fairly unpredictable process. Applied as an extension of phase-music principles,
Series is hard to beat. In more conventional applications, it's hard to fathom.
Preliminary experimentation proved Series great for generating harmonically static background textures, which could conceivably be
transposed in a sequencer to fit into conventional harmonic progressions. In all likelihood, though, there's a great deal more
lurking in this program's code; it's very good at coming up with musical materials that, while controlled, are not obvious from the
input.
- Stochastic Algorithym Composer (SAC).
-
This is the closest the package comes to giving you instant music: All you do is enter a tempo, 16 pitches, four MIDI channels, a set
of 16 note-or rest-values for each channel, and a choice of staccato or legato for each channel. The program will happily grind out
random lines in four voices 'til the cows come home. Depending on your choices, the results will either be cacophonous or
Utra-New-Age, though it's possible to shoot something in between.
There's actually a little more to it, enough to give you some interesting kinds of control over the output. First, you can set a
transpose value for each voice. This widens the output's harmonic scope somewhat, since otherwise each voice has the same choice
of 16 pitches. A controlled,though fairly static, harmonic scheme can be constructed by paying attention to all possible intervalIic
combinations of the pitch entries together with their transpositions. In addition, you can trap a certain length of the resulting
music say, four bars to be looped. If you choose to permute that looped portion, it will exchange a couple of pitches and note values
every few repetitions. This kind of controlled repetition and variation can give you some pretty coherent output.
Like the other two programs, SAC derives much of its value from the user's sense of discrimination. While it's a trivial matter to
arrive at values that produce good sounding results, making music out of them requires, at the very least, careful input selection.
The next step is to separate the wheat from the chaff by block editing in a sequencer, and after that to edit the individual lines.
The program's tendency toward harmonic stasis, for example, can only be overcome by waiting until it produces something that sounds
like a chord change, snipping that section out, and discarding the rest; then it might be desirabie to fix the voiceleading.
Nonetheless, for those who are satisfied with the raw results and they can be very pleasant - there's virtually no work involved in
obtaining them, and they can be customized somewhat to the uses's tastes.
Given the user's potential role in editing the products of these three programs, you might want to look into a program tailor-made
for doing just that: Dr. T,s new version of the Keyboard Controlled Sequencer
