Dice game music

KOAN PRO. Sseyo +44 1344 712005 Pounds 139.95 + VAT Windows diskettes (16 and 32 bit versions included).

GENERATIVE MUSIC 1 BRIAN ENO Sseyo +44 1344 712005 Pounds 38.29 + VAT Windows diskette (16 and 32 bit versions included).

There are many computer programs that both musical beginners and experienced musicians can use to create more or less original pieces of music. Some generate music with little or no intervention from the user, others act as tool kits. Yet others can be operated in both of these ways. But no music generating program seems to have made quite the same impact in the national press as Koan, a program developed by the British company Sseyo.

Koan runs on PCs equipped with a sound card, and at a pinch on a PowerMac under SoftWindows. No musical keyboard is need, though one can be accommodated. A few minutes after installing the program, the purchaser can take one of the pre-existing templates, optionally adjust a few parameters, and leave Koan to generate new pieces of ambient music. (Indeed the parameters may be adjusted to make changes while Koan works.) For those wishing to use the program more in the manner of a tool kit, one can dispense with templates, and experiment by trial and error to produce something from scratch.

No doubt some of the recent media interest in Koan has been due to Brian Eno, who has said that he would have been proud to have composed some of the pieces produced by Sseyo using Koan. Eno went on to use Koan himself to generate all the pieces for his new album Generative Music 1. The album comes as a 3.5-inch floppy disk of 12 tracks that will effectively never play the same twice - a feature of some importance to Eno. The variation that occurs on each playing is a built-in feature of the underlying technique used by Koan. Koan allows the composer to choose values for some 150 parameters governing aspects of pitch, rhythm, phrase length, harmony, tempo, key and so forth, but rather than forcing exhaustive specification of the exact result, the program encourages the composer to specify a range of choices and adjust their relative likelihood of occurrence. Within such constraints, Koan makes its choices randomly every time it runs. Eno claims that such programs represent the beginning of a third category of music: not live music or recorded music, but a new kind of musical collaboration between people and computers. Eno believes that in time the potential uniqueness of each performance associated with programs like Koan will become important to the general listening public.

Although Koan introduces no new techniques of principle into the area of computer generated music, and is limited in the genres to which it is well adapted, it has other significant strengths which may make a difference to the way all such programs are developed and treated in the future, and may pioneer the extension of their use to new kinds of application and new user communities.

In order to get a clearer picture of Koan's strengths and weaknesses, a good place to start is with the kind of music it produces. The program is designed particularly for composing in one specific genre, namely ambient music. This term was coined by Brian Eno to describe pieces such as his "Music for Airports". Eno uses the term to describe pieces where the creator works with pre-specified materials, but in unforeseen interactions and combinations.

Koan is very well suited to this kind of music, but it is probably not very appropriate for working on conventional tonal music or pieces with intricate harmonic structures. Having said that, it would be hard to put clear-cut limits on what it can and cannot do well. Koan can be used to produce good examples of styles such as techno and some kinds of minimalism, but in effect these may be viewed as variants or close relations of ambient. If you do not like ambient music then this limitation of Koan is obviously a problem. On the other hand, the limitation can be seen as intelligent focus on strength on the part of the designers. Ambient, with its local order but relative lack of highly interconnected middle and higher-level organisation is the genre that perhaps most obviously matches the strengths of the probabilistic techniques used.

While many music generation systems, including Koan, have had serious musical effort put into obvious structural elements, such as melody, rhythm and harmony, Koan also pays meticulous attention to providing ways of harnessing subtle expressive factors such as micro-variations of pitch and dynamics; shades of timbre, use of samples, sustain, portamento, reverberation, spatial placement and so forth. The demonstration piece "Forest" is particularly successful in making inspired use of these features. This capability allows Koan to produce breathing, organic, delicately shaded and nuanced textures, noticeably more alive that those produced by most music generation programs. This is one of the strongest features of the program. Of course the results will limited by the quality of the sound card used - the results on a cheap sound card will be disappointing.

The underlying probabilistic techniques used by Koan are not at all new, being similar in principle to those used in the first computer composed music, generally agreed to be Lejaren Hiller and Leonard Isaacson's 1957 Illiac Suite (a Quartet for Strings), with antecedents in systems such as Mozart's "Dice Game K294d". Mozart's "Dice Game" can be used to generate minuets using dice and strips of music cunningly prepared by Mozart to work in any combination. Such probabalistic techniques do not have the scope and structural reach of those used in, say, artificial intelligence influenced systems such as Peter Desain and Henkjan Honing's Music Loco or Peter Stones' Symbolic Composer. Neither do they have the gesturally explicit real-time control of such very different programs as Music Mouse, m, Upbeat, Different Drummer, Oval Tune, Harmony Space or Harmony Grid. Nor is Koan designed to make explicit the musical ideas that underlie successful pieces produced using it.

Koan does have one key feature that no previous music generation program has had in such measure, namely a thorough, imaginative approach to new kinds of use in a variety of contexts. This may have implications beyond the obvious.

There are several key innovations in this area. To start off with, Koan is sold as two different programs; Koan Pro, a program for creating and playing ambient music, and Koan Plus, a program (looking and sounding much like a little CD player application) for only the playback of pieces. This is the program used, under licence, on the floppy disk used to deliver Brian Eno's album. Koan music files are now appearing on some Web sites. Two alternative programs for playing these files - the Koan Web helper application (for most browsers) and Koan Plugin (for Netscape 2.0) - are available free at http://www.sseyo.com. Note that files used by the Koan player do not contain any potentially bulky sound or MIDI data, but just very concise data on the values of parameters generated by the creation program. This has interesting implications, as music could be sent much more quickly and easily over the Internet in this form than sound or MIDI files. The clear distinction made by Koan between producer and consumer applications facilitates the possibility of various market niches developing in compositions produced in this way. The Internet, and especially the World Wide Web, provide an obvious forum for such trading to take place.

The makers, Sseyo, discuss the possibilities of purchasers using ambient music produced by Koan Pro on Web pages, video and films, in screen savers, restaurants lifts and public places. For unchaperoned live music generation in public, a licence is required. For just about anything else, including, say, soundtracking films, no licence is needed. Sseyo provides an application programming interface (API) to allow Koan players to be controlled from application programs.

The copyright issues have to be considered in several layers, as variously affect pieces, recordings and MIDI files of particular performances. Careful thought has been put into this and documented in the manual, although there remain some murky areas in law. The Koan Pro User's Guide discusses how composers might choose to retain or relinquish their various layers of copyright, with the software offering facilities for authors to record such decisions in the files they produce. These measures will help reassure both buyers anxious to get royalty-free music, and composers anxious to protect their work.

The mouse, keyboard and graphical interface have evidently had a lot of work put into them, giving Koan Pro a very professional feel. At the same time, the connections between the various parameters and the way they interact are not always clear and the best way of empowering the user to control and visualise these connections has probably not been found yet. There may be other ways of organising the interface to address these problems.

Does Koan have any implications for the teaching of music, particularly in higher education? Certainly any such program can be used for stimulating ideas, but more systematic approaches are possible.

Programs that generate music are probably better suited for education when the musical knowledge they incorporate is represented in richer formalisms than probability tables, or when their design exploits the psychology of music perception or advanced human computer interface design with specific educational purposes in mind.

However, Jeanne Bamberger of MIT has proposed simple ideas for using more or less any programmable music generator as an educational resource by the simple expedient of encouraging students to imagine the result they wish, or expect, to achieve before they trying it and listening to the result. Students are encouraged to try out the effect of each change and then reflect on the gap between intention and perception. Due to the careful attention to all matters ambient, and on the control of micro variations in expression in particular, Koan might be a useful resource for encouraging reflection on these subjects. Other educational approaches discussed by Diana Laurillard of the Open University, in her book Rethinking University Teaching could also be applied.

Koan specialises in ambient music, and it is not clear how popular ambient music will be. Koan introduces no fundamentally new composition techniques, but the attention to detail, especially in matters of nuance, and the radical rethink of how to disseminate and apply such programs in new contexts make it unique.

It is just possible that Koan, in combination with the current flowering of all things connected with the Net, may create a series of new niches for computer generated music of several kinds. My own hope is that this does happen, and that programs using a variety of techniques and specialising in other styles and facilities will help Koan develop new communities with an active involvement in music.

Simon Holland is a lecturer at the Open University. His research areas include human computer interaction, artificial intelligence, music and education. Email: s.holland @open.ac.