in competence: algorithms

Anyone who knows my work can guess that in competence was substantially made with my slippery chicken algorithmic composition software. A lot of free-form editing and development of score materials were made in Dorico. The 5th-order Ambisonics sound files were made, collated, and mixed in Reaper using IEM’s plugins. My image-to-sound algorithms were the main new synthesis development for the project. Featured alongside this, right at the beginning, is a significant and highly processed recording of a speech from Richard II (some of it through my trusty old workhorse the TC Electronic Fireworx), as detailed below, along with my own reading of this. See the programme notes and my post on the recording for the general background to the work and note that the complete score is freely available online.

form, Shakespeare’s lines, recording proportions

This can also be represented by the following screen shot of an auto-generated reaper file with markers (‘L’ stands for Line). This formed the meta-structural orientation blueprint for the project. Click the image to get the full-resolution graphic:

snippets

Contrary to the general principle of slippery-chicken, where rhythm sequences and pitches have an abstracted relationship, in competence, makes significant use of a new object-oriented Common Lisp class called snippets (and related classes snippet and snippet-vars). snippets are objects which define small musical ideas consisting merely of notes, i.e. specific pitches and rhythms. When a particular snippet is in use, then only the rhythms and pitches defined in the snippet are deployed, though they may well be transposed over the course of the main music-generation algorithm (sequence). This is specified by a transposition curve given to the sequence method. Variations of snippets are expected to be defined and these are automatically deployed by the sequence method.

A good dose of snippet repetition is fundamental to the approach and software logic, as is a movement through the given snippets driven by the procession algorithm or good old fibonacci-transitions. Snippets don’t have to fill whole bars but they may contain barline delineations. They will in any case be combined into bars in order to create the whole sequence and for incorporation into a slippery-chicken object, if so desired.

The snippets classes and sequence algorithm was used in pieces 2 and 3, but also in 5 and 6. The latter are essentially rhythmically augmented and elaborated versions of the the snippet structures from no. 3.

A good example of this algorithm in action is the “Virtual Super-Marlies” sound file which begins 10a. farewell: improvisation:

hockets

Audible as repeated lower notes in the above snippets example is a form of hocketing. This technique is fundamental to in competence and is particularly highlighted in piece 3 from Letter B onwards (page 28 of the score PDF). Essentially, an existing slippery-chicken part is duplicated but instead of copying notes it is the rests that are filled in with newly-selected notes or chords, i.e. all notes are turned into rests and vice-versa:

In this particular case, the sopranino sax part is generated by sequenced snippets and the piano is created by hocketing, selecting chords from an algorithm based on the analysis, sequencing, and elaboration/transformation of piano chords originally improvised by Karin Schistek.

layers

This new, quite simply algorithm makes its entrance in the very last piece of in competence (11b. richard their king is dead). It produces layers (aka voices, parts) that repeat permutations of a list of rhythms in a variety of different augmentations. The given rhythms are central to the algorithm whereas the pitches produced are merely rising chromatic in the first instance, in order to clarify the structure. The actual pitch structure is assumed to be obtained or sequenced post hoc via another method. The layers algorithm takes as its main arguments:

1. a list of an arbitrary number of rthm symbols e.g. (e s e q) = 1/8th note, 1/16th, 1/8th, 1/4;
2. a list of scalers to apply (circularly) to the layers; this can be of any length and the default list consists of the prime numbers 7, 11, 13, 17, 19, 23
3. the number of layers/voices to produce;
4. the number of times the scaled rhythms should be repeated in each layer; implicit here is that the faster layers will end before the slower layers.

There is also the possibility to thin out the notes in the layers via an envelope, with Y values ranging from 0 (all notes will be forced to rests) to 10 (all notes sound). This is not done randomly, rather deterministically, via an activity-levels object.

foreign material

Pieces 7 & 8 come from algorithms developed ten years ago for a new precision mechanism is governing the city (guitar and mandolin). This itself was based on a quite radical departure of the hyperboles algorithm I was developing around the same time. It involved the rhythm-chains algorithm that was not itself used in hyperbole: there was a point during the development cycle that things went haywire and something I wasn’t particularly aiming at appeared. I love these moments (usually). I couldn’t use it for the piece I was working on but I made a mental note and kept the code for some future piece. It was only when I was working on hocketing for in competence that I remembered all this. It had a similar feel and so I developed it in the context of in competence for the Trio Abstrakt instrumentation. It’s quite nice, for me at least, to listen and compare the in competence version to the 2013 original, called whoops my algorithm slipped, as rendered here for synthesised flute and piano: