George Lewis (b. 1952)



Voyager (1993)
Recorded February 6 1993 in Alfred Hertz Hall, UC-Berkeley

Engineered by Laetitia Sonami
George Lewis - Trombone
Roscoe Mitchell - Alto and Soprano Saxophone

1. Voyager Duo 1

2. Voyager Duo 2

3. Voyager Duo 3

4. Voyager Duo 4

5. Voyager Duo 5

6. Voyager Duo 6

7. Voyager Duo 7

8. Voyager Duo 8

9. Home Coming


Voyager (the program) analyzes aspects of an improvisor's performance in real time, using that analysis to guide an automatic composing program that generates complex responses to the musician's playing. This implies a certain independence of action, and indeed, the program exhibits generative behavior independent of the human performer. The system is not an instrument, and therefore cannot be controlled by the performer.

If you want to hear the program play in a certain way, you figure out how what you want to hear is organized, and you start playing that way yourself. Te result is two parallel streams of music generation- that of the computer and that of the human, each informed by the other's music-all improvisational, subject-subject model of discourse, rather than stimulus/response

The software that performs on this recording considers at least thirty different musical parameters as it works on accepting input and generating output. Te input information is passed to a number of analytic processes that deposit their outputs into a block of variables that amount to the state of the input at a given moment. Volume, sounding duration, octave, register, interval width, pitches used, volume range, frequency of silence, and articulation are checked and averaged over time.

All of this is passed to the music-generating processes. Many of the same parameters found in the input are passed to the output, such as frequency of silence, volume, sounding duration, octave, register, interval width, volume range, and articulation.

Other important musical choices are generated internally via random numbers. These processes provide much of the "personality" of the system, and include melody and harmony, orchestration, ornamentation, pacing, transposition, rhythm, and internal behavior options, such as whether and how to react to input, or how quickly to change parameter and which should be changed. In the absence of all input, all needed parameters are generated using random numbers.

One hears a sort of orchestral music, an improvised concerto with an improvising orchestra. I do love the sound of the traditional orchestral instruments, and I am not looking for novel timbres via the computer, except as such might arise from the instrumental combinations that the software comes up with. Thus the work is not about synthesis in any way. I am happy to profit from the work of those who are interested in such things, and I hope that through the sale of high-quality synthesizers and samplers, they are profiting as well.

What the work is about is what improvisation is about: interaction and behavior as carriers for meaning. On this view, notes, timbres, melodies, durations, and the like are not ends in themselves. Embedded in them is a more complex, indirect, powerful signal that we must train ourselves to detect.

The Voyager project was started during 1985-86, while I was a composer-in-residence at the Studio voor Elektro-Instrumentale Muziek in Amsterdam. I would like to thank the entire past and present staff of STEIM, including Michel Waisvisz, Nico Bes, Gabrie Lantinga, and Misha Mengelberg, for giving me the time and freedom to work on this music.

Early versions of Voyager were originally developed for and programmed on the Yamaha CX-5 computer by George E. Lewis, using a special version of FIG-Forth developed for that underpowered machine by STEIM's technical director, Joel Ryan. The first complete versions of Voyager for the more powerful Atari ST series of computers, were finished in 1987 in New York City, as usual under extreme budgetary constraints. This version was written in Mach 2 Forth-83. Both of these early versions used a simple real-time scheduler of my own design.

The versions of the Voyager software heard on this recording was designed and written by George E. Lewis, using Forthmacs, a version of Forth-83 written by Mitch Bradley, and MacMox 1.4, a real-time scheduler and MIDI parsing program built by Danile J. Scheidt. Some Forth words designed by Phil Burk for use with the Hierarchical Music Specification Language were also used in this version of Voyager; I hope he won't mind. Software for the session ran on a single Apple Macintosh Powerbook; both the 145 and 180 models were successfully employed. A Sony lavalier microphone was used to capture the sound of the instruments. An IVL model 4000 Pitchrider was used to convert this audio to a form that the software could process. The instruments played by the software was an Emu Proteus/2 sample player and synthesizer. The software has 12 instrumental channels at its disposal at any one time. Any one of the 45 Proteus "instruments" (timbres) that I chose for the recording can be selected by any instrumental channel.
- George Lewis




UbuWeb Sound | UbuWeb

PennSound | CENTRO | EPC | WFMU