I wrote the following research paper about generative and algorithmic processes in digital music and production for Konrad Strauss' A201 course. This course occurred during the Fall of 2008, during my sophomore year at the Indiana University Jacobs School of Music.
In a culture defined by preciseness, conciseness, and perfection, the concept of non-linearity is a radical one. As electronic hardware becomes increasingly more computer-centralized, the ability to calculate dense operations under severe multitasking environments has taken priority over previous technological focuses. In the music world, this factor has become increasingly impactful in the spectra of composition and production. These days, it seems that a great piece of music, whether it be a standalone musical work, or a soundtrack to go along with the latest blockbuster film, video game, or television program, must conform to a range of precise production requirements. While this has undoubtedly produced a countless number of musical compositions that have perfectly fit the needs of both the producer and the consumer, it has also led many musicians, connoisseurs, producers, and executives alike to feel artistically alienated from the commercial music world. Indeed, organic process is a difficult characteristic to preserve in music so dedicated to linearity.
One such problem emerging from the issue of overly linear production via cutting-edge technology is that it can cause the resultant music to feel exceedingly bland. As legendary musician/composer/producer Brian Eno put it in the October 2001 issue of Wired Magazine, one cannot “do anything interesting with cutting-edge technology except not make it cutting-edge.” Like Mr. Eno, many have grown disdainful of commercialized musical establishments, which promote composition and production through the use of software like Acid, Logic, and Cubase – applications that allow the user to focus on an infinitely fine and detailed level. Because of the growing perfectionist trend in the commercial music world, a new musical philosophy is emerging, which in fact borrows concepts from a rather ancient technique. Generative music, which is more contemporarily known as non-linear audio, is a term popularized by Mr. Eno, used to describe music that is created by a system. (2) A technique expanding in popularity astronomically, it is a method being used by all ends of the musical spectrum to counter the perceived lacking in organic quality of contemporary music.
Musicians who implement non-linear audio into their works typically do so through the use of computer technology. (2) Instead of composing a work of electronic music that sounds exactly the same each time it is played, non-linear composers will establish a musical system defined by musical rules. These systems will usually be made up of a series of complex algorithms, which will produce similar – but not the same – results each time it is performed, as the algorithms can easily be modified by randomization or other nonrecurring data. As an analogy, imagine a door entrance to a shop with chimes hung from it. As one enters the shop, the chimes will move as the door moves, causing the chimes to strike one another. The resultant sound will be similar in character each time the door is opened and closed. However, the effect can be altered to a certain extent if one opens the door more lightly or powerfully than normal, as the chimes are directly connected, and reactant to, the door’s movement. Opened more lightly, the door will cause the chimes to be affected less, and more powerfully, the chimes will be affected more.
Similarly, non-linear musicians and composers strive to create their own musical systems for their works. Contemporary examples include video game soundtrack scorers, whose compositions change characteristics as certain user inputs occur; DSP patch programmers, whose applications affect audio based on algorithmic specifications instead of user-input; and the thriving community of algorithmic composers, whose works often include a combination of algorithmic computer code with generative properties and user input that alters those algorithms. They all share common ground, however: they modify and compose music by creating contexts, frameworks, and limitations.
Among the first truly generative contemporary pieces was Brian Eno’s 1975 magnum opus, Discreet Music. For this piece, Eno provides four instrumental passages recorded at a very slow tracking speed, and leaves it up to the performer to decide what the playback speed ought to be, and how to spool the tape reels (which measure many yards in length). The generative property comes from how these tracks are played back; the phasing and tonal properties of the piece will widely vary depending upon the decisions the performer must make. (3)
In the seventies, Eno prided himself most on the fact that works like Discreet Music are never performed the same way twice. However, as his works gained more and more notoriety in the music world, other artists began to incorporate their own ideas into this radical composition method, causing non-linear composition and production to provide an even larger range of possibilities than just mere differences in meter and phasing. In the early 2000s, a time where limitless editing potential was being provided by Pro Tools digital audio workstations, many were beginning to realize the dynamic power generative composition possessed. As engineers and entrepreneurs began innovating in the areas of real-time input and user interaction, composers and producers began naturally incorporating non-linear properties into everything from DSP plugins to video game scores. This new use of non-linear audio is most frequently given the title, adaptive music/production, in order to distinguish it from standalone works of generative music. (4)
The real significance behind adaptive production is the way it has allowed both producers and end users to merge the artistic depth of non-linearity with the recursive and interactive properties of user input. Before, composers like Eno were limited in that the composer and/or performers were the only people able to influence the way that their generative pieces grew and progressed. With the invention of adaptive music came the potential for the audience – the end user – to have just as much impact on the music as the composer. In the most practical of implementations, this spells out greatly expanded potential for dynamic interaction between user input and systemic output.
In the video game industry, non-linear techniques such as adaptive music have been used to highly diversify musical output during gameplay. Bungie Studios, which is responsible for the hit game series Halo, was among the first video game producers to implement horizontal re-sequencing and vertical re-orchestration in their products. Horizontal re-sequencing involves re-shuffling a series of composed loops depending upon user interaction within the game, while vertical re-orchestration involves literally changing the mix of a given loop in real-time depending upon the player’s movement within the game world. (6) Both processes require extravagant coding techniques, derived from complex mathematical algorithms, in order to most fully obtain a realistic-yet-immersive effect. Since Bungie utilized this technique, it has become nearly commonplace among all major game developing studios. By taking advantage of the latest that computational power has to offer, game developers have been able to further legitimize the in-game experience, by immersing players in a sound world all their own that they helped create.
Another rather unique example of non-linear audio in the video game industry comes from Nintendo’s Electroplankton, designed by media artist Toshio Iwai for the Nintendo DS portable game system. The game allows players to interact with animated plankton (each, a different “instrument”) and create music through one of ten “plankton interfaces” (types of score environments). By combining different types of plankton, and utilizing them in various patterns, users can create multi-hierarchal pieces of music. (6) There’s even an ingenious “audience mode” which algorithmically generates musical performances for the player to watch evolve, without even having to interact with the game system. Iwai claims that applications like Electroplankton are the future of adaptive music in video games. (6) Granted, the idea is quite provocative; his creation is perhaps the first entertainment product to ever be marketed as a generative music tool. To echo Mr. Iwai's assertions, the public and media reception of his product have been mostly positive. Nintendo Power magazine found the game so unique, that they classified it under a genre all its own: “touchable media art.” (6)
As can be seen by the sudden and powerful effect that the video game industry has had on non-linear music and production, the world of generative music is comprised of many various areas of interest. As commercial innovators like Bungie Studios and Toshio Iwai push ahead in the interactive gaming field, there at the same time exists a continuing interest in algorithmic composition and performance in the realm of contemporary art music. The fastest-growing collection of composers who pursue such interests reside, surprisingly, more in the field of software design. Computer music programming languages have quickly boomed in population among classical music composers, game soundtrack designers, and even neuroscientists who utilize generative music to study the sensory effects such works have on the human brain. (8) The most widely-used applications are SuperCollider and ChucK – both, interestingly, open source applications. SuperCollider is maintained by the Computer Music department at Wesleyan University in Middletown, CT, while ChucK is maintained by the Computer Science department at Princeton University; both share an abundance of enthusiastic contributing developers. (7) The reasoning behind throwing so much monetary and intellectual support behind these applications and developers becomes readily apparent when one observes the unique characteristics of these languages. Both languages are designed specifically for generating, calculating, and performing algorithmic processes, while at the same time serving demanding audio synthesis functions. Unlike applications like Reason and Acid, SuperCollider and ChucK do not contain a user interface unless the user wishes to build one himself. So, the obvious shortcoming to these languages is the initial learning curve. However, with none of the work done for the user, the user in fact can perform, control, and specify an infinitely broader range of tasks. (7)
While these programs are rarely known to the typical music listener, they are actually quite popular among many music and production circles, and have indeed been used by the most well-known and gifted of electronic musicians – everybody from Peter Gabriel to Aphex Twin and beyond. Their use is prevalent in the classical realm as well, as many of Iannis Xenakis's stochastic synthesis pieces have been created posthumously by means of SuperCollider and ChucK. During his time, analog audio devices limited Xenakis greatly due to their inability to deal with sounds of radically short durations. (7)
The list of contemporary innovators in the field of computer music programming languages is growing rapidly. One of the most prevalent individuals advancing the study of this ripe field is Nick Collins, who wrote the notorious BBCut library for the SuperCollider language. BBCut took music producers by storm with its non-linear methods of generating and cutting beat-driven samples. Instead of having to cut up drum beats manually, producers can use the BBCut syntax in SuperCollider to automatically render a beat map of the given sample, and then cut up the sample in real-time to give a consistent beat - containing specified characteristics, but lacking in linearity. His work has been immensely useful in live performance settings as well, since the BBCut syntax allows performers to produce these same drum beats in real time during performances, adding a heightened sense of organicism which used to consistently lack in live electronic music events. Because of his library's popularity, BBCut has been ported to many other platforms, including Max/MSP and pD, Csound, and even standalone VST plugins. The introduction of BBCut2, which came with a VST version upon release, catapulted its popularity to nearly mainstream levels, as many Pro Tools producers were introduced to the library for the first time. Presently, the BBCut2 library can be heard utilized by a number of popular musical outfits, including Battles, Radiohead, and Sigur Ros. Only in the last two to three years has the music world begun to witness generative properties featured in popular music lacking film or video game utilizations. The rapid spike in this particular library's popularity has certainly come to signify the impact generative plugins and applications are beginning to have on the commercial music world. (1)
In addition to the compositional non-linear adaptivity being witnessed in commercial music circles, generative properties are now becoming rather commonplace among recording engineers as well. The non-linear aspects of today's cutting edge VST plugins for popular audio editing programs such as Pro Tools have greatly broadened the realm of possibilities for engineers to tweak and improve their recordings. Similar to the dilemma composers face, recording engineers have to utilize electronic improvements tastefully, lest their work seem dull or overproduced. The latest VST plugins coded with non-linear properties aim to give engineers the ability to more precisely define their sound, while also maintaining an organic quality.
The desired effects that non-linear VST applications aim to achieve are particularly visible among the selection of equipment simulator programs currently available. These plugins are designed to simulate a wide variety of audio devices such as vintage tube amplifiers, preamplifiers, and delays, with the intent of recreating the actual hardware's behavior, instead of merely trying to imitate the desired sonic qualities such hardware possessed. Older plugins, which do not utilize this newly-adapted design technique, do not obtain nearly the same level of quality and clarity that non-linear simulators do, mostly due to the philosophy applied to its intended usage. Linear simulation applications indeed suffered many undesirable flaws because of their inability to actually replicate the physical properties of the intended audio hardware. Non-linear plugin designers remedied this problem by intending for their software to be used in a much different way; instead of simply devising a series of frequency, dynamic, and timing modifications which were meant to simulate the sound of the equipment, non-linear plugins are meant to simulate the actual device's circuitry. (10) This means that these simulation plugins take into account the variable properties circuits naturally possess, such as how the frequency and amplitude of an incoming signal affects the way which the internal components will operate. While non-linear and adaptive VST plugins are still priced at the higher end of the market, it has recently become much more feasible for the typical recording engineer to utilize the beautiful and natural processes that non-linear audio can achieve. (11)
Ironically, non-linear DSP has been just as helpful in removing the signs of vintage audio equipment as it has been at replicating them. As a continually growing number of analog audio recordings are being restored for digital formats, the need for more accurate and efficient noise removal has risen as well. Before non-linear techniques became typically utilized in remastering studios, the act of noise removal was an utterly painstaking process. Many engineers understood the implications of using non-linear filtering as a noise removal technique, but computer hardware simply could not deliver the processing power required to do so. With the rapid increases in computing power witnessed over the past 15 years, however, the ability to utilize numerous non-linear filtering theories has become progressively more possible. There are two techniques most often used today. The first method is called deconvolution – the act of removing undesirable convolution in recordings, such as tape hiss and analog floor noise. Deconvolution involves implementing mathematically-randomized chance through an algorithm that will ultimately deliver the equation
f * g = h
where f is the convoluted signal, g is the signal convoluting f, and h is the recovered signal. It is the object of deconvolution applications to determine the g signal, in order to remove it and ultimately make f = h. (10) The difficulty with utilizing this technique was that computers were unable to separate noise signals from musical signals, since often times the frequencies of both overlapped. However, when computing power caught up to human innovation in the mid 1990s, computer engineers were able to devise a method to separate the two signals by having applications calculate the amplitude differentials between them in real time. When this became possible, computers could then remove noise signals from musical signals with a relative amount of accuracy. (11)
Though deconvolution is still the most often-utilized method of noise removal in remastering studios, homomorphic filtering is a new technique becoming widely popular. While it was originally conceived at MIT in the 1960s, the technique has just recently been championed by computer engineers for a variety of applications. The method involves implementing the same scanning algorithm used by deconvolution programs, but instead of removing the detected noise signals, homomorphic filters actually largely increase the gain of the noise before using a reverse logarithm to remove it altogether. (10) Because the noise is much louder than the rest of the signal, homomorphic software can much easier detect noise that will interfere with the musical recording if removed. It is for this reason that many engineers have praised the technique for its ability to better preserve the original signal's fidelity. Whichever technique is ultimately deployed, though, it is undeniable that non-linear applications have immensely increased the power audio engineers have over recordings today.
The fact that music outfits worldwide, both compositionally and commercially, have begun to embrace the concept of non-linearity to a noteworthy degree, is quite revealing of its contemporary progression. Most fundamentally, it shows that the original reasons behind the development of generative music have become rather moot. While it was at first a way for composers to supersede the musical mainstream, it has now expanded and developed into an artistic language all its own, which has indeed intercepted much attention from more common performance and production methods. However, to claim that generative techniques have become outdated, or something to rebel against in its own respect, would be improperly short-cited. The ways that the sub-genre of adaptive music and production, for example, have impacted the music world are nothing short of spectacular – both in a sense of artistic merit and production value. Furthermore, the community of algorithmic composers is a strong and maturing organization (5), which will be seen in history as one of the prevalent musical processes of this era and beyond. The fact that their ideas have pervaded more commercial musical bodies so successfully demonstrates the inherent variety of obstacles that non-linear audio can overcome. More importantly, however, the success of non-linearity in the mainstream means that the perpetual notion of organicism in music is all but dead; in fact, quite the opposite is true. With the further continuation of non-linear DSP techniques, and the nearly universal implementation of adaptivity in interactive media, musicians, producers, and end-users alike will all be treated to a much brighter artistic future. It is indeed wonderful to see that the computing power of the 21st century will be utilized in such a humanistic manner.
Bibliography
(1) Collins/McLean/Rohruber/Ward. 2003. Live Coding in Laptop Performance. MIT Press.
(2) Dorin, A. 2001. Generative processes and the electronic arts. Organised Sound, 6 (1): 47-53.
(3) Eno, B. 1996. Generative Music. http://www.inmotionmagazine.com/eno1.html.
(4) Essl, K. 2002. Generative Music. http://www.essl.at/bibliogr/generative-music.html.
(5) Levitin, D. 2007. This is Your Brain on Music: The Science of a Human Obsession. Dutton Adult.
(6) Lieberman, David 2006. Game Enhanced Music Manuscript. In GRAPHITE '06: Proceedings of the 4th International Conference on Computer Graphics and Interactive Techniques in Australasia and South East Asia, ACM Press, Melbourne, Australia, 245 - 250. http://portal.acm.org/citation.cfm?id=1174472
(7) McCartney, J. 2002. Rethinking the Computer Music Language: SuperCollider. MIT Press.
(8) Sacks, O. 2007. Musicophilia: Tales of Music and the Brain. Knopf.
(9) Shachtman, N. 2001. New Eno Music Gets 'Generative'. Wired Magazine, 10 (2001).
(10) Smith, W. 2007. The Scientist's and Engineer's Guide to Digital Signal Processing: Chapter 22: Audio Processing. http://www.dspguide.com/ch22/7.htm
(11) Winkler, T. 1998. Composing Interactive Music. Cambridge, Massachusetts: MIT Press.
No comments:
Post a Comment