Generative Musical Improvisation For Participants With All Kinds of Musical Backgrounds
Media Lab, University of Art and Design Helsinki, Finland
This paper presents a generative musical improvisation process. During the moment of playing, there is a contact between the sound created by the performer and the sound generated by the interactive system. The improvisation turns into a continuous composition as long as a performer takes an active role in the performance. An important goal of the performance is to provide a situation where a participant of the performance can experience a free collective musical improvisation. It is a generative musical improvisation environment for participants with all kinds of musical backgrounds.
In the history of music sequencing, from the appearance of the first automatic instrument Aeoles , in the early second century BC, up to this time, there have always been new ideas coming forth on creating and developing new musical instruments, styles and theories. There is a cross influence relation between technology and existing music forms in any timepiece. Although technology does not invent a new form of music, it opens up new possibilities for musicians to expend their abilities to offer new musical values. Non-traditional and computer based interactive systems have been potential environments for these kind of possibilities in music . Early examples of computer generated music compositions are limited by the requirement of computer technology, however they encouraged musicians to compose computer programs to create models for their compositions. Computers have been affording composers a high level of control over the evolution and the variation of organizing sounds.
Computers, by framing a ground for generative music, free the composers from a note-by-note composition. Without worrying about the actual note and the existing musical structures, computers allow composers to concentrate more on the process. This creates an identical characteristic of generative music, which concerns itself more with the process rather than a predefined outcome.
Developing techniques for transforming the process into randomness or into a concept of chance, constructed different models in generative music compositions. Statistics have been used to derive musical composition from mathematics and logic. Xenakis, a major figure in the area of algorithmic composition, developed stochastic technique with his compositions . His approach was based on random generation and probability theory, which is used to generate material in a number of ways. Statistical analysis has been used as a modeling technique in most of the algorithmic music compositions. The interactive system of the performance is using random walk, drunk walk probability method to generate notes in its algorithmic structures. In this method, information is linked together in a series of situations based on the probability of the sequences. Random walk, drunk walk, has similar probability structure as Markov chain process. Markov chains based on the sequences of random variables in which the future variable is determined by the present but is independent of the way in which the present state occurred from its predecessors . Probability is the core statement of the decisions in Markov chains sequences.
During a collective musical improvisation there is a continues communication between musicians. The interactive system that runs in the performance creates this communication with performer on the listen and respond ground. This two-way interaction supports the improvisation’s beautiful and ever-changing identity and brings the unexpected beauty on the stage.
When it comes to music, the word improvise can be defined as performing without a musical score, or producing sound in a performance without preparation. It is a way of playing that gives unexpected results. Karlheinz Essl, Austrian composer, improviser and performer, describes the real time in which improvisation takes place, is passing by where one has to follow a certain way which might have been thought about before or which turns out to be negotiable during the improvisation . One must be continuously conscious of references; what has happened before, how can this be developed further? At the same time, during a collective musical improvisation there is a continues activity, such as exploring new sequence of sounds and listening consciously . Responding immediately and spontaneously to one other’s playing creates the process of communication in free collective improvisation. Within theses immediate responses there is also conscious awareness of how the sound is connected to other musician’s sound. Free collective improvisation, which is based on formless conversation, is a dialogue between musicians. No one knows where the conversation will go. The outcome of a free collective improvisation is unpredictable. This makes it possible to link performance practice of a free improvisation to any generative work process. Free improvisation concerns itself more with creating the dialogue between musicians rather than predefined performance outcome.
The task of the interactive system that runs during the generative musical improvisation performance is to create this dialogue in real time between a performer and the system. In this context performer’s actions, musical outputs, affect the interactive systems outputs and also interactive system’s outputs affect performer’s outputs. Performer’s musical actions are the needed input information for the system to generate notes, rhythms, to assign tempo changes and also to suggest scales to be followed during the improvisation. Performer improvises and the interactive system responds by improvising. It is a collective improvisation performance.
Involving the work of computer in the decision making process makes it possible to explore more about the machine aesthetics in this performance. Human-computer interaction is the main building block of this improvisation process.
Figure1. Interactive system of the From Me to Us installation was developed further to be a generative musical improvisation tool for this performance .
“Everything around us can be represented and understood through numbers.” Max Cohen 
The musicians of ancient Greek built their musical systems upon the theoretical applications of numbers and various mathematical properties derived from nature . These properties were the formalisms, or algorithms. These early algorithms did not involve the musician to be entirely removed from the decision making process of the composition. However musical systems of intervals and modes which are based on the concept of the “Music of the Spheres” are undoubtedly important historically in music for its leaning towards formal non-human processes.
The Webster’s College Dictionary defines algorithm as a set of rules for solving a problem in a finite number of steps, in order to find the greatest common divisor. A sequence of steps designed for programming a computer to solve a specific problem . In musical applications, algorithms may be thought of as procedures that test potential compositional material for its suitability within the given context. With this definition of algorithm, an algorithmic composition can be described as the application of a well-defined algorithm to the process of composing music .
During the past decades various interactive systems have been implemented in computer generated music applications. These implementations are constructed on their suitable algorithmic music theories. Music theories have their individual identities in relation with their time and application systems, however there are similarities and relations between each other as well. Their evaluation may even be built on the earlier efforts of other implementation. Robert Rowe identifies the applied systems’ similarities and relations by developing a classification of interactive systems . The computational methods of classification, which are in touch with this performance, are the interactive computer response methods. Interactive system is constructed by using Pure Data environment.
Pure Data is a freeware graphical programming environment for real-time audio and graphical processing. Miller Puckette has been actively developing Pure Data computer music environment. Pure Data was written for multiplatform from the beginning. It is open source software that can be downloaded, free for any use and development . It is a work in progress.
Electric guitar that gives two outputs has been chosen as a performer’s instrument in order to communicate with the system. Analog output directly goes through chorus and over-drive effect processors. Second output goes through MIDI guitar interface where analog sounds are transformed into MIDI information and connected directly to the computer. By filtering and analyzing the musical information that comes from the performer, computer system runs three major decision making units. Tempo-rhythm generator, note generator and record-sample units.
Note generator is a part of a system that follows the player as well as a part of a response technique method. Performer defines the musical structure that will be followed during the performance, however the system also randomly proposes the certain scales. Randomness is the core statement of the decisions in the system’s response process. Performer’s musical inputs are analyzed in order to give immediate response to a response to create a free collective improvisation process.
The random outcomes in the performance are determined so that the generated notations will not be dependent on any predefined structure. Even though certain scales are followed, there is no predefined structure concerning the order of notes that the system generates. No two performances are identical. Following the scales with non-predefined notes creates the controlled random identity of the interactive system.
The performance provides a situation where a participant of the performance can experience a free collective musical improvisation. It is a generative musical improvisation environment for participants with all kinds of musical backgrounds. The performance brings together the ever-changing musical identity of a free improvisation and the work of computer in the generative process.
 Muro, Don and Purse, Bill. 2000. TI:ME Course 2A Advanced Sequencing, course material. Available from:
 Tahiroglu, M.Koray. From Me to Us, a computer generated music installation. In Proceedings of the Participatory Design Conference (PDC '04), Vol 2, Toronto, Canada, July 27-31, 2004, CPSR, P.O. Box 717, Palo Alto, CA 94302, p12, 2004.
 Fonseka, Joseph Rukshan 2000. Musical Agents. Thesis (Master of Science). Monash University. Available from:
 Franz, David M. 1998. Markov Chains as Tools for Jazz Improvisation Analysis. Thesis (Master of Science). Virginia Polytechnic Institute and State University. Available from:
 Essl, Karlheinz 2002. Improvisation on "Improvisation", ed. Hauser, Jack . Available from: http://www.essl.at/bibliogr/improvisation-e.html.
 Nunn, Tom. Wisdom of The Impulse. On the Nature of Musical Free Improvisation. 1998 (publ. by author), pdf version (IIMA) 2004.
 Pi . Film. Directed by Darren ARONOFSKY. USA, 1998
 Maurer, John A. 1999. A Brief History of Algorithmic Composition. Paper. Available from: http://ccrma-www.stanford.edu/~blackrse/algorithm.html.
 Jacob, Bruce L. Algorithmic Composition As A Model of Creativity. Organised Sound, vol. 1, no. 3; Special issue on algorithmic composition. Cambridge University Press, December 1996.
 Rowe, Robert. Interactive Music Systems : Machine Listening and Composing. Cambridge, Massachusetts: The MIT Press, 1993.
 Pure Data Community Site is available from : http://www.puredata.org/