Contents

Artistic Research 5
SUBJECT : Structural Time and Spatial Composition 5
Context 7
parametric composition 9
tessellation, resolution and scale 11
dimensions and projections 15
cinematic, simulation and dynamic 17

Implementation 21
choice, generative series 21
workflow 22

Bibliography 29
Artistic Research

SUBJECT : Structural Time and Spatial Composition

Mathematics, geometry and computation are possible inspiration sources and tools leading to an intellectual
piece of work. This art piece does not need to be accurate or cartesian at all. Formalized ideas are never-
theless often used for constructing forms (gestalt) and patterns 1 which can physically be materialized in
architectural design. Using another medium, such materialization can also result to musical scores or purely
sonic pieces.

Each medium brings a different reading and comprehension of a sin- 1

James Tenney and Larry Polansky.
Temporal gestalt perception in music.
gle piece; the correlation between them supports the apprehension of 1980
the work. A piece of music using other media can be apprehended in
the same way an architectural project could be presented 2 . Beyond the 2
Kas Oosterhuis. What exactly is
final sound itself, the work passes thru several contexts, inspirations nonstandard architecture? 2009

and its conception. For instance, the question of contextualization con-

cerns installations, theatre and dance pieces. But it also concerns vari- 3
Julie Steele and Noah Lliinsky. Beauti-
ous ways of representing the work using data visualizations 3 or even ful visualization. 2010
writing books for example.
Music scores are an interesting case of study particularly for pure elec- 4
John Grayson. Sound sculpture. 1975
tronic music. Both prescriptive and descriptive notations exist for this
kind of music too 4 . But their conception and use are fairly different
from scores for acoustic instrumental pieces. Beyond the fact it is also
a piece of art, the use of an electronic music score has no particular
order of reading. The simple traditional order starts from the concep-
tion, its interpretation, and the final listening. There are no such steps
because there is no need of instrumentalist interpretation 5 . Also all 5
Olivier Messiaen. The technique of my
musical language. 1956
levels, happening in the same "box", can occur at any time of the cre-
ative and listening process. Such scores can then be the cause and the
consequence of the music piece. We have seen many examples in the
60’s with graphic scores. Also today, audio-visual interactive pieces us-
ing interfaces set the player both as a listener and a composer. A huge
amount of work has been done in this direction in the past decades 6
Carsten Nicolai. Auto pilot. 2002
and many synesthesic pieces have risen from it 6 .

The project of this thesis concentrates in a slightly different direction.

The aim is to apply compositional processes both to space and time.
6

The score then becomes one entire part of the piece; it is the spatial 7
Michel Jacobs. The art composition.
composition of the piece 7 . It can be presented in many forms and 1926; and Graham Nerlich. The shape
of space, second edition. 1929
context: mostly a piece of music but also a sculpture, video, 3d print-
ing, paper... Although they are the piece of art, they still are called 8
James Gips. Shape grammars and
score because they can be read in any direction of space, in the same their uses. 1975; and Stephen Wolfram.
A new kind of science. 2002
way one would read text or an evolution of patterns 8 . It is not here
question to recreate another generalized writing for electronic music.
The link between sound and audio does not have to be proven in a 9
Richard E Cytowic. Synesthesia. 2002
cartesian way 9 . The reader or listener discovers a world by himself
while listening or reading evolutions. His perception and interpreta- Jean-Claude Risset. Rhythmic para-
10

tion would maybe have more significance than the actual audio-visual doxes and illusions. 1997
connections of the piece 10 .

The text has the tendency to be didactic and describe a true theory
in physics, psychoacoustics, musicology or aesthetics. It is rather one
contextualization of my work in the same way architects would present
one project with technical, social, historical and economic contexts. My
personal thoughts could possibly transport the public into a purely
abstract and unreal universe. This world is part of the art and would
eventually give one continuity of my work.

Olivier Pasquet - Berlin - July 2014

 7

Context

Below are mutual researches and consequently aesthetic orientations found in pieces presented in this doc-
ument.

Tours Creations
Pieces TTF AHI ALI NBP STV ASS JAM RIB HR8 R13

Pure sonic composition X X X X X X X X X X

Spatialization (pan, wfs, ambi) X – X X X X X – – –
Spatialization (env) – – – – X X X – – –
Generative X – X – – X – – – –
Visualization – – – X – – – X – –
Visual composition – – – – – – – – X X
Sculpture X – X – – – – – – –
Nano-composition – – – – – – – – – –
Staged composition – – – – – – – – X X

TTF: Tu Tiens sur tous les Fronts, AHI: Aucun Homme n’est une Isle, ALI: Aliados, NBP: Nuits Blanches Paris 2013,
STV: Steve V (King Different), ASS: A Script for Synthesis, JAM: Joyeux Animaux de la Misère, RIB: Rib, HR8: hr
8798, R13: r136a1

Research subjects in the table are set in a specific order; from the starting point with a traditional approach
to composition, to the final point of the research where composition is at its widest definition. "X" should
have the tendency to go down the table in a near future. That would mean I am reaching the radicality and
formal depth I am looking for. One difficulty and one question is beyond technical aspects; should the "X"
disappear from the top of the table ? Should I give up my current work or find a clever way to link both
aesthetics.

· Pure sonic composition is the most traditional technique consisting on

writing for all compositional dimensions such as timbre, rhythm, dy-
namics etc. There is no focus on only one of them but they all have an
important role. For example, Philippe Manoury uses grammar or rules Philippe Manoury. Les grammaires
11

to compose elements by elements 11 . The approach is one extension of musicales génératives. 2012
integral serialism.
· Spatialization (pan, wfs, ambi) adds spatialization to the writing. Sound 12
Matteo Meneghini. Stria, by john
displacement are composed as any other compositional dimension as chowning. 2003

John Chowning did in a simple but efficient way in Stria in 1977 12 .

There are many tools for composing movements of sounds. There are
also several techniques for properly perceiving sounds sources and
their existence in a specific space.
 8

· Spatialization (env) makes use of envelope spatialization in which sounds François Xavier Féron, Julien
13

are reduced to quantic entities close to the limits of perception and Boissinot, and Catherine Guastavino.
Upper limits of auditory rotational
understanding. For example, Emmanuel Nunes made brilliant use of motion perception. 2009
envelope spatialization although he was not really consider extending
it to its limits 13 .
· Generative shows pieces where some formal and symbolic generation 14
Tony Myatt. Processes and systems in
is used. The result tends to be an evolution of short quantic rhythmic computer music; from meta to matter.
2011
patterns 14 .
· Visualization is the line for pieces which are in some extend connected 15
Pierre Boulez. Das fruchtfeld. paul
to a visual part from analysis. I am not planning to work on this part klee. 1989

already explored by many other artists. 15

· Sculpture shows pieces which are not only sonic but also visual. Both 16
Valentina Croci. Techniques and
media are being read or listen as if the score was the piece itself. An technologies in morphogenetic design.
2010
important part of the composition processes is being done before the
choice of medium. The final piece is an audio/visual object. This is the
true architectural part 16 .
· Nano-composition shows pieces also using quantic rhythmic patterns 17
Eshel Ben Jacob. Bacteria harnessing
often close to the limits of perception and understanding. But they complexity. 2010; and IBM. A boy and
his atom, the world’s smallest movie.
are not always sound pieces. They are composed using synthetic biol- 2013
ogy and/or nanotechnologies 17 . Such pieces are composed integrally:
from a microscopic to a macroscopic scale. Time scale for sonic parti-
cles and pitches, space scale for quantic structures. These works can
be presented in the form of installations with or without sound.
· Staged composition are pieces using approaches described previously 18
Stamatia Neofytou-Georgiou. The
but, this time, with a confrontation to reality. For instance, the ques- semiotics of images in romeo castel-
lucci’s theatre. 2010
tion of constraints from the stage, other people, and how to perform
an electronic piece rises again. Interrogations about audience, social
positions, interfaces between thought and physicality are back. Those
confrontations can then be tools for more societal subjects as it is often
the case 18 .
 9

parametric composition

Figure 1: ’jtol.bach.humanize’
randomizes rhythmic trees on a
grid

I have written a Max library for rhythmic trees with the help of JT Rinker. I started this collection of tools
before starting the doctorate but it is still under progress. It is a library dedicated to real-time pattern gen-
eration and can be applied to architectural design, music, dance; everything requiring constructed evolution
in space and time. It deals with multi-scaling and multi-dimensions where rhythm is considered to be a
skeleton onto everything else is attached (pitches, params...).

A pulse stream is deducted from another stream of discrete events. Pul-

sation (felt or not felt) induces a metrical structure which can be or-
ganized then represented using a particular data structure. Each ratio
before a new parenthesis level is the total duration for what is inside
this parenthesis. For instance, simple musical tuplets would be ex-
pressed in the following way:
( 1 ( 1 1 1 ) ) => ( 1/3 1/3 1/3 ) => ( 0.333 0.333 0.333 ) (Σ 1)

Rhythms are often represented with rationals because they can easily
be apprehended by humans; this is only one reason. This representa-
Figure 2: Meter induction
tion has its limits from a mathematical or computational points of view
when formal problems occur. Also, the quantic and fractal disconti-
nuity of meter can sometimes generate very complicated situations for
improvised, non western or oral transmitted music 19 but this is not
19
Gilbert Nouno. Suivi de tempo
the case when music is played by a machine. An algorithm would have appliqué aux musiques improvisées.
problems defining and finding onsets in a pulse stream but it would 2008
be much easier to accurately position events, accents etc.

If time is segmented using bars, each bar and its contained discrete
events can be set on a circle [3], a spiral or hyper-spiral. Rotation rep-
resents time. The duration of a bar can be either set by the sum of
10

what it contains. It can also be independent and scale all inside dura-
tions according to a fixed duration. This second approach is used for
rhythmic tree representations and can sometimes be very useful when
durations are fixed by factors external to the inside of the bar.

Figure 3: The duration of a bar

can be represented by the cir-
cumference of a circle for in-
stance

We are here talking about limits, continuity, integral calculus and

moreover differential equations for complex situations like composi-
tion using several parameters. For instance, problems like homorhythm
can make use of such mathematical, geometrical tools. A function is
considered as an infinity of points infinitely close to each-other at a mi-
croscopic scale. It is indeed relatively easier to find one common onset
from several rhythms using a function rather discrete points connected
to each other. In the same way quantification or swing techniques can
use similar intellectual tools 20 ; either a musical and empirical ap-
proach or a more formal approach using discrete or continuous rep-
resentations. The question to go from discrete to continuous and vise 20
Carlos Agon, Gerard Assayag, Joshua
versa involves musical decisions according to musical needs. It is a Fineberg, and Camilo Rueda. Kant: a
question of mapping; in a similar way we connect physical sensors to critique of pure quantification. 1994
musical parameters.

Instead of composing using pulses, it is then possible to compose using

"curves". The general term "curves" is here used as a generic geomet-
ric or articulative term. These "curves" could for example be functions
(bijective or not) or parametric equations. They are described with
21
Dmitri Tymoczko. A geometry
equations and parameters 21 . I call this parametric composition because of music. 2011; and Iannis Xenakis.
it can have deep esthetic musical implications as well as the revolu- Formalized music. 1992
tionary piece Metastaseis from Iannis Xenakis.

Connections with structuralism in mathematics can be seen. The ap- 22

The concept is closed to the ENP
proach to jtol is more orientated towards a discrete "new complexity" notation front end in PWGL or om:tree
in Open Music
rather than continuity 22 . A tree representation is indeed similar to
theoretic rhytmic techniques from Brian Ferneyhough or the music
23
Mika Kuuskankare and Mikael
from Richard Devine who both make intense use of nested irregular tu- Laurson. Expressive notation package.
plets and combinatorial processes 23 . The idea of complexity interests me 2004
more from the point of view of perception rather than its direct techni-
cal construction. Only the result count, so random or stochastic processes
 11

are part of it. Also, the notion of continuum from set theory concep-
tually and perceptively makes links between discrete and continuous
rhythm 24 . Visual rhythmic layers in the piece ??hr 8798, described ear-
lier, are so dense one could not literally understand anything; only the 24
György Ligeti’s sound piece Contin-
overall and perceptive understandings are still consistent. This consis- uum (1968) or Casey Reas’ visual piece
tency maintains a continuity. In other words, discrete time onsets are pfft (2014) are good examples here.
so close they perceptively build a continuous evolution of texture.

tessellation, resolution and scale

Figure 4: A deep zoom inside

my piece Kaspar V (2012) made
with deconstructive caustics pat-
terns using super high resolu-
tion imagery

Discrete visual information can be seen as raster graphics. Continuous information, based on mathematical
expressions, can be seen as vector graphics. There is here a close connection with computer music: para-
metric music would be vectorized and pulse music rasterized. Vectorization leads to the generalization and
organization of information; it eventually leads to data compression. Many artists used connections between
raster (rhythm, moirés, lines etc) and vector (interpolations etc) to create synesthetic connections; Rijochi
Ikeda or Ryoichi Kurokawa are interesting examples in that field.

In typography, a point is the smallest whole unit of measure, being 25

Sinan Bokesoy and Jean Baptiste
a subdivision of the larger pica 25 . These points can be positioned or Thiebaut. An approach to visualization
of complex event data for generating
classified onto a plane but they do not need to touch one another. It is sonic structures. 2008; and Martin
also the case with traditional music scores with a note, or a group of Solomon. The art of typography. 1986
notes. Such symbols are not always connected together. Those symbols
can sometimes take the form of nested or tiled data structures; tuplets
12

for instance. It is nevertheless different with signal, sound. Events 26

Sigrid Block Philippe Veenendaal
are positioned onto a continuous and unswerving timeline. Sound Diederik Williams Chris Adriaenssens.
Shell structures for architecture. 2014;
must always fit into a continuous timeline because time is entropic and and Johann Rafelski and Berndt Müller.
cannot physically be stopped. There cannot be "holes" into a signal; The structured vacuum. 1985
only connected silences or discontinuities. It is different with symbolic
writing in musical fields or architectural design where it is possible to
have "jumps" 26 . Tectonics would be interesting to endeavor.

Tessellation is consequently necessary for signals such as sound. On

the other side, it is not an obligation for symbols such as scores. But
everything depends how signal and symbol are distinguished. Also,
the reader of symbols has the freedom to read in any direction and
dimension; graphs are therefore useful for scores. Only staffs can give
a direction for time in the case of traditional music notation.

A score describes how will eventually be the content of the final piece.
But it is also a piece of work by itself. It is then a composition for Figure 5: Vectorized Penrose
another composition. It is then possible to conceive a graph, with its Tiling
representation of information, as a piece by itself. It is then possible
to consider the evolution of a graph into a defined space rather simply
considering it as changes of states.

We saw there is an apparent continuum between "raster" (granular)

and "vector" (continuous) symbols when perception plays with reso-
lution 27 . For instance, zooming inside a Penrose tiling [5] allows the
reader to decrypt the information from another perspective with dif-
ferent details or a different temporality. That zooming can be discrete 27
Lindsay MacDonald. The limits of
or continuous. But once again here, continuity comes from the limits resolution. 2010; Bob Sheil. Ad - high
definition. 2014; and Reginald Bain.
of a quantized (granular) universe with minimal elements acknowl- Algorithmic composition quantum
edged as being beyond human understanding or perception. This can mechanics & the musical domain. 1970
be called a quantic continuum.

Many music theories make use of a quantic approach 28 . Sound is 28

Douglas JE Nunn, Alan Purvis, and
considered as granular even at its smallest scale. Granular syntheses Peter D Manning. Acoustic quanta.
1996; and Curtis Roads. Microsound.
utilize sound grains which can be considered as quanta. It is similar 2001
with grain from visuals (pixels, grain of picture etc). Sonic grains can
be composed onto a timeline organized in pitch, dynamics or many
other sound descriptors. Textures are created using tiling techniques
like concatenation of chosen grains together, overlapping, quantizing
onto a time grid, random positioning etc. The technique was widely
used in the 1990’s 29 . It quickly becomes very musical because time is
involved. It is now largely employed for music information retrieval or 29
The architect Michael Hansmeyer or
gabor signal processing. Pure compositional techniques are used when photographer Andreas Gursky play
grains are considered as being long. The technique becomes a true with scale and multiplicity. Each micro-
element is a single system connected
synthesis when they are short enough and reach limits of perception, to others and create something on a
or understanding. Some composers use set theories to organize grains macro-scale.
 13

30 .Some others create patterns in which the final texture is made of 30

Forte Allen. The structure of atonal
both grains and their carrier, envelope or grid for instance. Alike for music. 1977; and Paul Riker. The
serialism of milton babbitt. 2010
clothes which prints are also sometimes dependent on textile itself 31 .

On one hand, synthesis techniques can be involved in signal processing. 31

Yohji Yamamoto often concentrates on
On the other hand, compositional techniques can rather be part of sym- the feel and touch of the clothing. The
bolic processing. I explained earlier there can then be a quantic con- early digital artist Otto Beckmann uses
tv pixels or laser dots as visible carrier.
tinuum between both processes. This continuum can be represented
by a graph with a nested structure. Quanta are joined together by
rules of inference. These rules can for instance be a bayesian network 32
Edoardo Acotto and Moreno An-
dreatta. Between mind and mathemat-
or a grammar network 32 . The overall is an organization containing ics. different kinds of computational
smaller sets of other organizations [6]. Each set, or cell, smaller or representations of music. 2012
with the same size, is considered as a scale.

A change of scale is an iteration in the compositional process 33 . We 33

Przemyslaw Prusinkiewicz and
talk about computation where time is unreeled from calculation. Aristid Lindenmayer. The algorithmic
beauty of plants. 1990

Figure 6: multi-scalar and multi-

order organization of informa-
tion. Hidden markov models
are an exception because it only
utilizes probabilities. In our
case, pattern cells can be gener-
ated by anything and are related
with others with rules. Mini-
mal patterns when zoomed at
the very bottom of the scale are
quanta(e).

For instance, fractal elements occur when rules are all the same ac-
cross scales. So fractals are exceptions because they always have self-
similarity or scale invariance between scales. Stephen Wolfram’s cellu-
lar automata or formal grammar Lindenmayer system are interesting
procedural generators within cells but they rarely totally fulfill the
conditions to generate an entire piece [8].

Scale Relativity is a theory of space-time initially developed by Lau- 34

Laurent Nottale. The theory of scale
rent Nottale 34 . Erwin Schrödinger’s work is also involved. It is an ex- relativity. 2009

tension of the special and general relativity applied to physical scales

(time, length etc) and aims to unify both relativities; a little like string
theory. It is used to predict the location of exoplanets for example.

The Planck scale is applied to mass, time and length. It is the small-
est observable possible scale not only because it maybe is the actual
14

Figure 7: The grey section corre-

sponds to what is understood by
humans. Physics breaks down
beyond the lines because the
model is still not accurate. There
is a suspect comparability with
the understanding of audio (dif-
ferent from limits of perception).
The reason comes from nature
and the way perception is hard-
coded in the human brain.

smallest scale but also because current laws of physics cannot con-
ceive further. This limit will probably be pushed away with another
"understanding" of the world. There is an alikeness with audio; sound
or music [7]. Beyond perception limits, there are boundaries for the
brain to understand the structure or characteristics of an audio signal.
It takes time to apprehend a intensity as much as it needs size to fit
mass. There is some kind of "incompressibility of energy". If the Plank
scale were maybe applied to audio grains, the quantic continuum would
then be a closed space. The reason of similitude is not actually linked
with physics. It is rather linked with something more general; it comes 35
Noam Chomsky. Syntactic structure.
from nature and the way perception is hardcoded in the human brain 1957
35 .

Figure 8: My piece 8799c (2013)

is not made with cellular au-
tomata but it involves an evolu-
tion of patterns using computa-
tion. Zooming in or out creates
interferences (moiré with pixels
from the screen). Density of dots
creates grey levels with a zoom-
out.
 15

Figure 9: Mesh of a minimal

surface made with the rhino 3d
software.

dimensions and projections

As described earlier, a graph is both a score and a visual piece itself. It is then not only a visualization
but also an evolutive form defined by a coordinate system and parameters. It is a parametric composition
as I defined earlier. The notion of parametric equation is generalized to surfaces, manifolds and algebraic
varieties of higher dimension. For example, a strait line in a 2d space is equivalent to a continuous change
of one dimension compared to the other. This phenomenon is similar to signals like frequency versus
time. It is different in the case of symbols where frequency and time are naturally de-correlated. They
are artificially linked together by an ordering structure build by the composer. The piece ends up being
this organization. This structure is then a multidimensional object in a space which cartesian coordinates
correspond to compositional parameters 36 . The objects are made of form, or gestalt, which can also change
over time. Several versions of a same object would just make a series of objects that could then be visualized
in a movie. It could also become a series of several music pieces. I am very interested by series because it
raises questions in relation with industrialization of music.

These forms can be constructed using various kinds of generative al- 36

Linda Dalrymple Henderson. The
fourth dimension and non-euclidean
gorithms one can find in music composition or architectural design 37 . geometry in modern art. 1983
Morphogenesis (biological process that causes an organism to develop
its shape) is for instance a great source of inspiration. John Frazer for-
malized reptilian structures as a tiling system. Also, the famous Karl
Sims was a pioneer on artificial life and composed evolving creatures
using genetic algorithms 38 .

Unfortunately, very few composers use such inspiration and instead 37

George Stiny and James Gips. Shape
grammars and the generative spec-
mostly use basic generators of form. One of the reason comes from ification of painting and sculpture.
cultural background and lack of open-mindedness of institutions. An- 1969
other reason comes from the very strong difficulty to associate gen- 38
John Frazer. An evolutionary archi-
tecture. 1995; and Karl Sims. Evolving
erative concepts with a musical interest. Strangely enough, using a virtual creatures. 1994
16

singular generative system ends with a result either too complex or

too simple to apprehend. It is a question of "good" ratio; harmony
between complexity and simplicity. It is also a main issue concerning
synesthesia; artistic (constrained) choices and concessions have to be
made between the triangle "generator" & "audio" & "visual". This sub-
ject is about mapping and remains one of the main artistic question
raised in this thesis.

Forms can then be projected onto a coordinate system. They can also
be projected to another form like a plane or a sphere for instance. A
projection can be seen as a reduction of dimension which is a reduc- Figure 10: Dots in 3d are pro-
tion of parameters for music. The object does not have to belong to jected onto a plane, the viewer.
an euclidian space; space deformation could deform the entire object The more perpendicular is the
hence its projections onto the coordinate system, hence its parameters. perspective angle of aligned
Projection also happens when a reader watches a visual score. In that dots, the closer they are from
case, the projection, point of view, can be set to a plane in the same each other. They eventually
way 3d softwares make use of virtual cameras. Positioning a camera draw a continuous line. A part
plane depends on what has to be shown from the object. Here also, it is of my piece Kaspar I (2010)
an artificial correlation between the abstract understanding of reading is voluntary shown here in a
and the pure visual aesthetics. It is an artistic choice. small size. But it is vectorized
and can be zoomed.
We have seen one piece is one object. But since it is generative, there
can be series. Each piece of a same series do not need to have the same
duration or tessiture. The object is transformed and only morphology 39
Robert Alan Dorgan. Music in archi-
remains. Here again, geometric and morphologic computations can tecture / architecture in music. 1990;
and Godfried T Toussaint. Compu-
be used for transformations 39 . Parameters can be used for changing tational geometry and morphology.
forms so we are still talking about parametric composition. For instance, 1986
we saw that symbols are naturally de-correlated from signals and that
only an artificial organization, composition, is connecting them. It is
easily possible to scramble coordinate systems and totally get another
result from a same form [11]. Algebraic geometry concerns the de-
40
Christof Migone. Volume of confine-
ment and infinity. 2003; and Georges
scription of properties of geometric structures using algebraic expres- Legendre. Ad - mathematics of space.
sions. It involves the study of "position" with topos and the study of 2011
form with topology 40 .

Figure 11: simple orthogonal ro-

tation of graph
 17

Figure 12: parabola and cubic

curve in a projective space. In
this example, the initial space
has infinite ends. The spheri-
cal one on which curves are pro-
jected is a limited space. Per-
mutation is normally a discreet
process: folding generates dis-
continuities. A projection onto a
sphere can rather be seen as con-
tinuous permutations.

More generally, music visualizations and processing often use forms

and geometry to better apprehend information in a visual way. Angle
of views, projections and dimensions changes are used a lot to trans-
pose that information into a better usable state and then use more
adapted tools. A simple example are chords or rhythms represented
or processed onto a circle [13]. A circle has been chosen because of the
relative cyclicality of musical scale or rhythms (modulo). This is also a
closed shape allowing to build and compare polygons.

Figure 13: simple permuta-

tions, shiftings, of one constant
rhythm in an eight beat bar

cinematic, simulation and dynamic

During the nineteenth century, Helmholtz studied the relationship between musical harmony and the human
perceptual apparatus 41 . His discoveries can guide an investigation into the relationship between music and
visual art. Helmholtz concluded that we appreciate the geometric progression in sound frequencies because
our ears seem to produce these overtones even in the absence of their physical presence. Time factor is
undoubtedly critical when reading, watching or listening information. Order (direction or linearity) and
speed are an important factor in the deduction of incoming signals.

The reader of those signals has a belief of understanding. This faith 41

Hermann Helmholtz. On the sensa-
tions of tone as a physiological basis for
drives him along what he thinks being a particular process. That pro-
the theory of music. 1885
cess does not have to be the one that actually generated the piece;
the reader can eventually consider whatever he wants in the way he
18

Figure 14: snapshots of a rhyth-

mic score of mine (2011). Note
the composed position of small
balls and their densities. The
piece can be read onto paper or
as a movie. Notice the impor-
tance of definition and scaling in
this piece.

wants. Temporality of reading is nevertheless forced with moving im-

ages pieces, traditional scores or even music. A timeline drives de-
duction in these special cases. The notion of memories then plays an
important role in inference. This is strongly the case with music even
though temporality can sometimes be considered differently; installa-
tion works for instance.

If the reader is free to interpret any incoming information, he can be-

Figure 15: bubble chamber;
lieve in actual scores but also in any other traces lefts by natural and
electrically charged particles
unartistic processes; a "score" from nature [15]. He can also consider
leave traces after collision.
data visualization or any other synesthetic object as a score although it
is not always the case. Level of complexity also has an important role
in the predictability of events or patterns. This is one of the reason why
randomness is widely used for composing seemingly complex struc-
tures. Complexity is again here an important area for experimentation.

Figure 16: 8799d a+b+c=a*b*c

(2014) triptych piece on pvc. It is
basically a 2d matrix of random
sets. Minimalism is on one hand
at its simplest and most possible
flatness. On another hand, max-
imalism is at its best; the most
incompressible pattern. One can
Time can then be unrolled when someone reads a score. The score only compare between the three
can then become a music using another media but using the same in- canvas; the medium and its con-
ferences. The score can also be dynamic and change over time. Time figuration becomes then as im-
portant as the content itself.
 19

becomes then just another dimension on which it is possible to project

parameters. Its structure then also becomes a tree structure with vari-
ous scales and rules as I described earlier.
Another time is consequently possible; a time eventually describing
physical time. Then another time describes the time that describes
physical time etc. There is no difference between time and space in that
case. The amount of times is the order of time structure. For instance,
a movie, already working with time, can see its timing changing with
various versions. Realtime music pieces always change performance
after performance because of their input and possible internal genera-
tive elements. Working on the time of a dynamic piece (time-based) is
working on a "meta piece". Some people call it generativity 42 . There Figure 17: This piece is based
is nevertheless a difference between time and space at the very end of on a single repetitive loop pat-
the creative chain; when time becomes real with its constraints in the tern translated in a plane by
physical world. This final step toward reality is actually the one that an autonomous multi-agent
confront a work to reality, to other people judgement. It is also useful system. It is again another min-
a feedback that can also sometimes be considered for corrections of the imalistic piece because only the
abstraction. This feedback loop is then another element in the entire path is read by visitors. It is
generative system and eventually give it life and stability. also a "complex" piece because
of its many and seemingly
It is also useful and inspiring to chose to separate time from space. random iterations.
Not only at the end of the creative process for physical movements 42
For his piece Solo, Karlheinz Stock-
like sound spatialization. I described earlier the possibility to com- hausen not only dictates the interpre-
tive processes but also rules the order
pose using topological spaces 43 or form themselves with topology
of the parts in a descriptive rather than
44 . It is also possible to use the missing link between symbolic form
fixed way.
and its surroundings by adding timed elements, by adding movements 43
Guerino Mazzola. The topos of music.
2002; and Georges Legendre. Ad -
and traces (again). This has very close links with iteration. Again here, mathematics of space. 2011
cellular automata, self-organizing systems or other computational ma- 44
Robert Cogan. Imaging sonic struc-
chines are involved. Spatial computing takes time into account in the ture. 1986
sense computation is space oriented 45 . 45
Louis Bigo. Représentations symbol-
iques musicales et calcul spatial. 2013;
Besides computational theory of mind, genomics and algorithmic trad- and Konrad Zuse. Calculating space.
ing, one of the ultimate work with computational complexity is the 1969
simulation of the universe. N-body simulations simulate a dynamical
system of particles. It studies processes of non-linear structure forma-
tion such as galaxy filaments and galaxy halos from the influence of
dark matter. A simulation imitates a real-world process or system over
time. It becomes composition when rules that govern it are initially
set by humans. In that sense, such multi-agent system is a rich spa-
tial composition. It would be interesting to change the rules without
caring about actual physics but rather artistic results. Douglas Hofs-
tadter’s Gödel, Escher, Bach is not about mathematics, art, and music
but rather about how cognition and thinking emerge from well-hidden Figure 18: Rather old but still
neurological mechanisms. Those mechanisms can nevertheless spring beautiful simulation of the
a totally imaginary universe materializing to series of "art" pieces. Universe from the Max-Planck-
Institut för Astrophysik in 2005.
Implementation

choice, generative series

Media Art is a generic term for art involving media technologies. The
subject is then often about the kind of media used by the piece itself.
This is the case here since the above poetical and theoretic explanation
is part of the piece of art and has an important role. I would rather
call this conceptual art because the concept is the personal unfinished
universe in which the next pieces should be taking place.

All of the above poetical and theoretic explanation is then applied to

rhythm onto score and sound. Igor Stravinsky said "There can be
no music without rhythm". I focus this work on rhythm because it
concentrates pieces to the purity of a structural skeleton. This does
not mean there is no sound at all but rather a minimalistic approach.
But sound somehow has to be part of rhythmic evolutions; little bit
as for traditional piano music. Rhythms on a nanoscopic scale can
generate macroscopic forms in the same way of nanotechnology.

There is not the intention to build a software that could be used by

anybody. It is rather the construction and organization of a workflow
allowing experimentation and composition inside the vision I describe
in the previous sections of this document. Also, there is not the wish to
generalize generative music nor the writing for electronic music. This
is more the work for musicologists and it is a huge risk that should
be avoided as much as possible. Such trap exists because of the re-
lation between computers and art. Choices have then to be done in
order to make the work the opposite of a generalization. Generaliza-
tion nevertheless exists for pieces inside a series and uniqueness only
applies to series as a single piece. I have the intention to build series
because they better relate their context. For instance, Pierre Soulages
is an interesting artist piece by piece but also for the work of his life-
time. Composing series brings the opportunity to present the piece
with a richer context. I want the context, and concept, to possibly be
shown as part of the piece itself. Far from being didactic or pedagogic,
they can just be inspiration sources or something completely fake but
 22

allowing the audience to "travel" and read the piece. I believe it is one
strength for architects who rarely only show a building without its so-
cial, historical and technical context. This strength must be used on
pure music for social and political reasons.

workflow

[19] summarizes how information should be organized and possibly

transformed according to the vision described in the previous parts of
this document. Three main directions describe the workflow the best:
· Scales are constructed using different rules. These scales are in relation
one to another using possibly different rules.
· Nested tree data structures are needed with the possibility to only gen-
Figure 19: summarize of how
erate or transform one layer at a time.
information should be orga-
· Form generators independently, or not, applied to each scale. nized and possibly transformed
according to the vision de-
It is legitimate in that case to use Music composition tools as well
scribed in the previous parts of
as Architecture composition tools. Tools are more or less adapted
this document.
 23

to the variety of scales; defined as "microscopic" or "macroscopic" for

convenience.
I use Rhinoceros and Grasshopper as the main system [20]. Rhinoceros
already has a tree data structure. It generates the "macroscopic" form
of the piece by placing onsets using a rich variety of advanced genera-
tive tools. On another side, Jtol, a homemade Max library, build nested
tree rhythms on a "microscopic" (sec) or "nanoscopic" scale (samples)
[21]. This influences both synthesis and the structure of the piece.
Those "microscopic" events are then hooked to the macroscopic onsets Figure 20: Grasshopper "def-
built by Rhino. All resulting events are linked to a series of "articu- inition" (external) i wrote to
lations" at any layer of the process. These "articulations" are kind of export to python (written in
multidimensional low-frequency oscillators or breakpoint functions. A C#...)
profound deep hack of AthenaCL is used for that. All that symbolic
data is then applied to sound engines and other systems for arrange-
ment and "improvement". CSound scores are generated and make a
very minimalistic music using simple waveforms and noises. It is the
same with SuperCollider. Ableton Live reads midi files containing
notes and controllers. Its API is also used to better controls virtual
and real synthesizers. It is then possible to compose with a less mini-
malistic and more conventional approach. The score part is built using
Processing for a minimalistic view; it cooks vertor graphics. A more
crafted version is drawn using Rhinoceros, Autodesk 3dsMax and v-
ray, a very powerful ray tracing plugin I know pretty well.
Figure 21: Jtol swing engine
Csound, AthenaCL and Rhino now all use Python [23]. This has be-
using Max
come the main language for this workflow. The choice of language is
also dependent from perennity and inter-software use. I have now be-
come very careful with programming becoming obsolete after several
! " #$ %& '% &! &" ($ $& )% "! ""

years. I use Max because part of this workflow could be used for live *+,+-./0-12.-.3+/+-456+7/817.9:;/

performance thru MaxforLive devices or just patches. The situation of a

! " #$ %& '% &! &" ($ $& )% "! ""

live performance with this offline system will undoubtedly come. *+,+-./0-12.-.3+/+-456+7/810@+-./0-A>=/:@=B

The workflow now seems to work properly, it is more than time to ! " #$ %& '% &! &" ($ $& )% "! ""

*+,+-./0-12.-.3+/+-456+7/81<.=>+?:+<+

make sounds and constraint the system even more in order to make it
a true musical instrument. Already interesting questions have risen : ! " #$ %& '% &! &" ($ $& )% "! ""

· The complex relation between the actual sound and its score. This is Figure 22: generated hacked
an ancestral and pure synesthesic question. How to make something AthenaCL functions.
visually readable, interesting and audible and understanding at the
same time. The easy and pretty way is to stay minimalistic. I have the
intention to use scaling to make it readable, minimalist and rich and
complex at the same time. The most obvious direction is the construc-
tion of spectrograms. They are a visual representation of the spectrum
of frequencies in a sound as they vary with time. Only 3 parameters
are visible on the score. This technique is the most efficient for the ap- Figure 23: Python is controlling
prehension of a piece in which rhythmic structure is important. Other the entire workflow.
 24

parameters than pitch, time and amplitude could be used; parameters

from a parametric composition could easily be seen 3 by 3. The next
step would be finding other ways to add and reduce (by projection)
dimensions.
· In the case of a simple spectrogram, events from Rhino are in a space
and not onto a timeline. We have seen a timeline could not have dis-
continuities and a symbolic space could have. Events can be placed
anywhere with Rhino... This rises a very deep question about hori-
zontality and the continuity of musical time. It is also a huge question
about verticality: how events are vertically fitting together. Figure 24: One polyphonic
· Another question is the way such piece has to be performed. The quanta generated with Python
fact a visual score can sometimes be provided for reading opens the inside Csound.
performance to a variety of formats such as installations etc. I have
the wish to compose series with two versions each. A-side would be
the "rich" version coming from Live. B-side would be the synchronized
skeleton coming from CSound. The audience could then compare both
pieces as for a diptych. Comes then the question of performance of
two pieces to be compared ! Again here, it is media art in the sense
it is about the medium itself, how a piece is performed and its social
reception.
· There are two versions: one "matrix" which is only the structure and
another one. The "matrix" sounds like typical computer music from the
1970-1980’s when academic computer music was not yet taken over by
mixed music in institutions. Pieces (pitched based...) with formalized
structures were fashionable. I am heading to this direction with nowa-
days’ technology: mostly calculation power, synthesis techniques and
speakers quality. I could add artifacts, a grain from the medium, such
as various kinds of background noises and short dynamics.

Figure 25: Extended phyllotaxis

spiral with Grasshopper and
Rhino. This example does not
have any micro-rhythms hooked
to events; so only one scale (or-
der one).

Figure 26: Out-coming synthe-

sized spectrogram with simple
sine-waves from CSound (the
buzz opcode is actually used).
 25

Figure 27: Other synthesized

spectrograms. One with only
one pitch all along the piece
but many changes of metric in
a microscopic scale. Another
one with a visible ternary form.
Again two others with phases.
None have any micro-rhythms
hooked to events yet. The other
version, coming from midi files,
and the API from Live, allow the
control of a wide variety of syn-
theses like virtual but also gear
synthesizers. Those gears could
be an opening to the question of
relation with a live control and
performance.
List of Figures

1 ’jtol.bach.humanize’ randomizes rhythmic trees on a grid 9

2 Meter induction 9
3 The duration of a bar can be represented by the circumference of
a circle for instance 10
4 A deep zoom inside my piece Kaspar V (2012) made with decon-
structive caustics patterns using super high resolution imagery 11
5 Vectorized Penrose Tiling 12
6 multi-scalar and multi-order organization of information. Hidden
markov models are an exception because it only utilizes proba-
bilities. In our case, pattern cells can be generated by anything
and are related with others with rules. Minimal patterns when
zoomed at the very bottom of the scale are quanta(e). 13
7 The grey section corresponds to what is understood by humans.
Physics breaks down beyond the lines because the model is still
not accurate. There is a suspect comparability with the under-
standing of audio (different from limits of perception). The rea-
son comes from nature and the way perception is hardcoded in
the human brain. 14
8 My piece 8799c (2013) is not made with cellular automata but it
involves an evolution of patterns using computation. Zooming in
or out creates interferences (moiré with pixels from the screen).
Density of dots creates grey levels with a zoom-out. 14
9 Mesh of a minimal surface made with the rhino 3d software. 15
10 Dots in 3d are projected onto a plane, the viewer. The more per-
pendicular is the perspective angle of aligned dots, the closer they
are from each other. They eventually draw a continuous line. A
part of my piece Kaspar I (2010) is voluntary shown here in a small
size. But it is vectorized and can be zoomed. 16
11 simple orthogonal rotation of graph 16
28

12 parabola and cubic curve in a projective space. In this example,

the initial space has infinite ends. The spherical one on which
curves are projected is a limited space. Permutation is normally
a discreet process: folding generates discontinuities. A projec-
tion onto a sphere can rather be seen as continuous permuta-
tions. 17
13 simple permutations, shiftings, of one constant rhythm in an eight
beat bar 17
14 snapshots of a rhythmic score of mine (2011). Note the composed
position of small balls and their densities. The piece can be read
onto paper or as a movie. Notice the importance of definition and
scaling in this piece. 18
15 bubble chamber; electrically charged particles leave traces after
collision. 18
16 8799d a+b+c=a*b*c (2014) triptych piece on pvc. It is basically a
2d matrix of random sets. Minimalism is on one hand at its sim-
plest and most possible flatness. On another hand, maximalism
is at its best; the most incompressible pattern. One can only com-
pare between the three canvas; the medium and its configuration
becomes then as important as the content itself. 18
17 This piece is based on a single repetitive loop pattern translated in
a plane by an autonomous multi-agent system. It is again another
minimalistic piece because only the path is read by visitors. It is
also a "complex" piece because of its many and seemingly random
iterations. 19
18 Rather old but still beautiful simulation of the Universe from the
Max-Planck-Institut för Astrophysik in 2005. 19

19 summarize of how information should be organized and possi-

bly transformed according to the vision described in the previous
parts of this document. 22
20 Grasshopper "definition" (external) i wrote to export to python
(written in C#...) 23
21 Jtol swing engine using Max 23
22 generated hacked AthenaCL functions. 23
23 Python is controlling the entire workflow. 23
24 One polyphonic quanta generated with Python inside Csound. 24
25 Extended phyllotaxis spiral with Grasshopper and Rhino. This
example does not have any micro-rhythms hooked to events; so
only one scale (order one). 24
26 Out-coming synthesized spectrogram with simple sine-waves from
CSound (the buzz opcode is actually used). 24
 29

27 Other synthesized spectrograms. One with only one pitch all

along the piece but many changes of metric in a microscopic scale.
Another one with a visible ternary form. Again two others with
phases. None have any micro-rhythms hooked to events yet. The
other version, coming from midi files, and the API from Live, al-
low the control of a wide variety of syntheses like virtual but also
gear synthesizers. Those gears could be an opening to the ques-
tion of relation with a live control and performance. 25
Bibliography

Edoardo Acotto and Moreno Andreatta. Between mind and math-

ematics. different kinds of computational representations of music.
2012.

Sigrid Block Philippe Veenendaal Diederik Williams Chris Adri-

aenssens. Shell structures for architecture. 2014.

Carlos Agon, Gerard Assayag, Joshua Fineberg, and Camilo Rueda.

Kant: a critique of pure quantification. 1994.

Forte Allen. The structure of atonal music. 1977.

Reginald Bain. Algorithmic composition quantum mechanics & the

musical domain. 1970.

Louis Bigo. Représentations symboliques musicales et calcul spatial.

2013.

Sinan Bokesoy and Jean Baptiste Thiebaut. An approach to visualiza-

tion of complex event data for generating sonic structures. 2008.

Pierre Boulez. Das fruchtfeld. paul klee. 1989.

Noam Chomsky. Syntactic structure. 1957.

Robert Cogan. Imaging sonic structure. 1986.

Valentina Croci. Techniques and technologies in morphogenetic de-

sign. 2010.

Richard E Cytowic. Synesthesia. 2002.

Linda Dalrymple Henderson. The fourth dimension and non-

euclidean geometry in modern art. 1983.

Robert Alan Dorgan. Music in architecture / architecture in music.

1990.

François Xavier Féron, Julien Boissinot, and Catherine Guastavino.

Upper limits of auditory rotational motion perception. 2009.
32

John Frazer. An evolutionary architecture. 1995.

James Gips. Shape grammars and their uses. 1975.

John Grayson. Sound sculpture. 1975.

Hermann Helmholtz. On the sensations of tone as a physiological

basis for the theory of music. 1885.

IBM. A boy and his atom, the world’s smallest movie. 2013.

Eshel Ben Jacob. Bacteria harnessing complexity. 2010.

Michel Jacobs. The art composition. 1926.

Mika Kuuskankare and Mikael Laurson. Expressive notation pack-

age. 2004.

Georges Legendre. Ad - mathematics of space. 2011.

Lindsay MacDonald. The limits of resolution. 2010.

Philippe Manoury. Les grammaires musicales génératives. 2012.

Guerino Mazzola. The topos of music. 2002.

Matteo Meneghini. Stria, by john chowning. 2003.

Olivier Messiaen. The technique of my musical language. 1956.

Christof Migone. Volume of confinement and infinity. 2003.

Tony Myatt. Processes and systems in computer music; from meta to

matter. 2011.

Stamatia Neofytou-Georgiou. The semiotics of images in romeo

castellucci’s theatre. 2010.

Graham Nerlich. The shape of space, second edition. 1929.

Carsten Nicolai. Auto pilot. 2002.

Laurent Nottale. The theory of scale relativity. 2009.

Gilbert Nouno. Suivi de tempo appliqué aux musiques improvisées.

2008.

Douglas JE Nunn, Alan Purvis, and Peter D Manning. Acoustic

quanta. 1996.

Kas Oosterhuis. What exactly is nonstandard architecture? 2009.

Przemyslaw Prusinkiewicz and Aristid Lindenmayer. The algorith-

mic beauty of plants. 1990.
 33

Johann Rafelski and Berndt Müller. The structured vacuum. 1985.

Paul Riker. The serialism of milton babbitt. 2010.

Jean-Claude Risset. Rhythmic paradoxes and illusions. 1997.

Curtis Roads. Microsound. 2001.

Bob Sheil. Ad - high definition. 2014.

Karl Sims. Evolving virtual creatures. 1994.

Martin Solomon. The art of typography. 1986.

Julie Steele and Noah Lliinsky. Beautiful visualization. 2010.

George Stiny and James Gips. Shape grammars and the generative
specification of painting and sculpture. 1969.

James Tenney and Larry Polansky. Temporal gestalt perception in

music. 1980.

Godfried T Toussaint. Computational geometry and morphology.

1986.

Dmitri Tymoczko. A geometry of music. 2011.

Stephen Wolfram. A new kind of science. 2002.

Iannis Xenakis. Formalized music. 1992.

Konrad Zuse. Calculating space. 1969.