Transmusic Background Information Method and Philosophy The TMF Format Music Media Issues Method and Philosophy The Isle Ex exhibits, while modest, can at least hope to broach some basic themes of a "transmusical project". The would-be transmusician hypothesizes that one may behold a diversity of autonomous dynamic structure through the common lens of music, that one may do so by applying a simple and consistent method, and that the resulting music can be fresh and interesting - both in itself, and as a new appreciation of the source system (though I'm not sure these two aspects can or should be separated). But let's begin at the beginning. Method In the technique developed to produce the Isle Ex examples, transmusic is generated by passing a series of unsigned integers[1] through a simple filter or musification map controlled by several parameters. In the case of the present rather primitive exhibits, only tone (MIDI key number) is mapped-to.[2] The renderings I've called multimaps were assembled by taking the results of several musifications using different unimaps (i.e., different parameter sets), and shuffling such monophonic "loops" around in series and in parallel, such as suited my fancy at the time. In making these simple multimaps I found myself following a few self-imposed "rules". Firstly, the loops I shuffled around were always exactly one musical period[3] in length. That is, the single-period loop was my atom of "transcomposition". I mixed and matched loops with various degrees of frolicsome abandon, but I never split them apart or otherwise tampered with them internally. A second rule I seem to have followed was to generate the melodic variety of loops within a particular multimap piece by manipulating either the N or TIV parameters, but generally not both.[4] Both rules, as should be obvious, are restrictive of my own role in the musical results. As it were, I purposely cranked up the autonomy dial on the transmusic machine, so as to let the source signal shine through (or, to avoid responsibility for the results, the cynic might say :) Transmusical Dawn? While I make no aesthetic claims for these exploratory efforts, and I'm not even sure what would be involved in doing so, I have little doubt that real musicians, as opposed to philosophically-minded tinklers such as myself, could go far with similar methods and sources. Serious transmusical systems would, of course, need to feature additional tone-mapping algorithms, and would map to other musical elements such as rhythm and dynamics. And such systems should be capable of realtime transmusical rendering. If such transmusical tools were to appear for cellular automata, for instance, one imagines that "bands" and "orchestras" of CAs might become a composition/performance medium in their own right, producing pieces whose aural and visual elements are organically-associated (that is, stemming from the same underlying complex autonomous dynamic) rather than "glued together" on the basis of ephemeral aesthetic prejudice. In this manner music may free itself from the ghetto of the ad hoc and post hoc, and ascend to the dignity of the propter hoc. The ancient Greeks contrasted techne with phusis - i.e., "that which is made" with "that which grows". Music has traditionally been regarded as belonging to the domain of techne. My transmusical experiments, by contrast, are best seen as part of a trend toward phusis in music. That is, increasingly, music will not be "made", but will "grow", or, to conceive of things in more basic terms, it will take much of its form from an independent source. Composing will become less like manufacturing, and more like shepherding, birdwatching, treeclimbing, cartography, swimming. Music will no longer be "ours" - and yet we will have lost nothing, and gained a whole world. In the words of a site visitor who had been attracted by the geyser musifications, as her epiphany hit during an email exchange: [...] couldn't you apply your transmusic filter to space noise? How 'bout weather patterns? Oh, or earthquakes! That would be cool. Geez, actually there are about half a million things. Your stomach growling. Immigration patterns. By Jove I think she's got it. And if transmusic is fueled by life itself, who can doubt that it's a "growth industry". Some Take-home Questions The transmusical endeavor raises a number of intriguing theoretical issues, many of which have not yet, as far as I know, been addressed anywhere. For example: What is "autonomy" in music? To view a piece of music not as the work of a particular human being, but rather as stemming from something independent which has been passed through an human "filter", would seem to involve a significant theoretical shift. But a traditional composer also has many "outside" influences. (Indeed, if we look carefully there may be nothing left "inside" - in dwelling on a particular human "creator" we perhaps make a fetish of a small node in a buzzing social dynamic.) In terms of its autonomous content, what if anything allows us to distinguish transmusic from other music? Assuming we can come up with a viable notion of autonomy, will we celebrate it in music the way we purport to do elsewhere? Or will we insist on keeping music on the plantation, torturing it in the music schools, binding its limbs in formulae, making it carry water for "genius". Perhaps music is more in need of a Moses than a Mozart. The possibility seems worth considering. How might we approach the question of the respective contributions to transmusic of the source system and the transmusician? (A corollary issue to that being, if you don't like the music, is there anyone you can blame? :) In considering this topic, keep in mind that if one micro-manages the transmusical parameters to an extreme, one can produce an arbitrary melody from a given input. It wouldn't do to lump such "pieces" in with the present ones, would it?[5] Related to the above question is the nagging worry that perhaps the methods described here on Isle Ex can produce music similar to our examples from essentially "random" data. I have heard others state something along this line quite confidently, but myself consider it an open question. An intriguing alternative possibility is that "musifiability" is a fairly sensitive gauge of hidden order. Clearly, this is a question which invites empirical investigation - any takers? Is there an "aesthetic" theory applicable to orchestrating transmusical parameters? I rather hope not - if there can be good and bad transmusicians, I'm in trouble. But seriously, though we had thought to escape the prevalent ad hominem fixation, we may instead be faced with a paradox wherein the human element, vis-à-vis transmusic, is skilled precisely in felicitous suppressions of its own influence. (But, come to think of it, that may still be our escape. Radical undecidability is effectively negative - "not to decide is to decide", and such a decision is definitely not in the affirmative.) If there is an aesthetics of transmusic, does it also have an ethics? Above, the rather colorful image of music-in-bondage was presented. But a declaration of musical independence really is less about the liberation of music per se, which in the final analysis is just a series of noises, than about allowing the most honored of such noises to be authorless, hence lacking in human ownership and control. As such we seem only to have encountered in the domain of music the familiar pattern of opposition to nature that is the hallmark of sedentary society (not "Western", as it is typically cast). That being the case, we find somewhat to our alarm that the issue of subjugation arises most pointedly within the practice of transmusic itself. For instance, supposing I were to transmusically coerce population data from an endangered sea turtle species into producing the theme from "Gilligan's Island" - ouch, talk about "shock art" :) These questions just scratch the surface of the subject, but you get the idea. Perhaps you'll even be the one to answer some of them, or at least ask them more definitively. The TMF Format The advanced transmusical rendering systems presaged above would need a "transmusic format" to specify a complete multimap composition/performance. That is, the "mixing" of unimap-rendered loops, which I formerly did manually in a sequencer, should be done by the musification software itself - or, to put it differently, the software must schedule parameter changes for multiple simultaneous musification "channels". So, for the sake of demonstration, I've developed a simple prototype of such a format. Such "TMF files" are in ASCII and thus suitable for editing manually. The format is pretty clunky, but it gets the job done, the proof of which is that I did actually use it to generate my most recent multimap pieces (i.e., the four multimap geyser-activity renderings). Assuming you've read the discussion of the musification process (see above explanation together with this parameter reference), the TMF format should, for the most part, be quite easy to decipher. So I'll limit myself to illuminating the darker corners with the following tips: The source series may be multidimensional, so I've used spreadsheet or database terminology to refer to its elements. Thus the series consists of a sequence of rows, and each member of a row resides in a specific column. For example, the input to the Botryoidal Spring multimap has two columns, viz., the eruption intervals and their corresponding durations. The parameter settings for each map or track definition carry over to any subsequent definitions in the same section, except where specifically overridden. Each of the TRACKS must reference one or more input columns, one reference for each musical element being mapped-to. In the present examples only tone is mapped-to, so each track references only one column. The 'R' in RMap stands for "rhythm", though 'Map' is presently something of a misnomer, as the current types take no input. (The name anticipates additional types that would use separate input columns to modulate note lengths and intervals.) The uniform type produces notes of equal length, one for each input value, whereas the glued type extends a note if the next converted tone has the same value. The \L command in the schedule causes the source to "loop" (i.e., the entire sequence is repeatedly input) until the end of the schedule is reached. This only makes sense for a finite input series, obviously. The SMEASURE field ('S' is for "schedule") specifies the number of input rows per schedule step, over the base length of the converted notes. The NMEASURE (for "notational") on the other hand is just the traditional time signature, which is encoded in the output (e.g., MIDI) but has no bearing on the musification process. Music Media Issues My musification software outputs MIDI, which is a music format rather than an audio format. The distinction is analogous to that between a musical score on the one hand, and a recording of a performance of that score on the other. (No one would confuse the two, and say things like "scores sound bad", yet the similarly confused "MIDIs sound bad" is not infrequently heard.) As a score format, MIDI is compact, flexible, and suitable for presenting compositional ideas, but poses serious difficulties when it comes to performance over the Web. Instrument timbres, volume responses, and so on, vary greatly from one MIDI synth to the next. (In the traditional realm, if a composer just handed a score to the audience and said "here, play it - I trust you", their mileage would probably vary too. They might even come up with the notion that "scores sound bad" :) There are also problems with some player software not playing all the MIDI channels. For instance, the "Eternal Recurrence" CA music piece uses 14 channels, and some Windows players don't play the upper 5. Also, a Mac I tested dropped channels from "PopCyclic" for unknown reasons. Now, I don't mean to lay all the performance shortcomings in the examples at the doorstep of irregularities at the client end. Far from it, the MIDI itself is raw, no-frills, untweaked. I gladly leave performance-level MIDI work to those more capable of it. Still, as Dante pointed out, even in hell there are preferable neighborhoods. So, in that spirit, I've recently rendered all the multimap pieces[6] from MIDI to digital audio. While these supplementary MP3s are crude[7], they are at least a known quantity. Notes Though there aren't yet any examples on Isle Ex, I've used the same system with floating-point sources such as chaotic maps, by simply using a prefilter to map them to unsigned integers. It doesn't really matter; the basic concept is the same, though modular mapping may be contraindicated. (Of course, technically, the geyser studies use a floating point source, since they involve measurements of a continuous variable. The source series were "prefiltered" before I ever saw them, by rounding to the nearest whole time unit.) As for other musical elements, some of my more recent efforts (viz., the geyser multimaps) introduce rhythmic variety by adding a second static way of handling rhythm, but ideally rhythm and other elements would also be handled by mapping schemes, operating on additional, simultaneous inputs. The distinction between musical period and source period arose in the context of my CA studies. For physical sources such as geysers, the musical period would normally just be the length of the input sequence. Certain contrived sets of musification parameters could introduce intra-sequence periodicity, but in musifying physical sources I have avoided such maps in the interests of fidelity. The Old Faithful Geyser musification is a unique case. The melodically distinct loops were rendered from four different input sequences of the same length, while the musification parameters remained unchanged. Hence, according to my tentative terminology, this rendering is a polyphonic unimap rather than a multimap, but the mix-and-match technique employed was the same as with multimaps. Interestingly, though, in such cases, while the contribution of the source system is no longer visible in the output, it is still present in masked form. It could be retrieved if we were to provide some unrelated input, a constant, say, to our over-programmed musifier. This sort of thing cries out for an analysis, not to mention possible use in producing highly novel pieces. (Program your musifier to produce a constant tone from a given input. Now input something else to that musifier. What do you get?) The MIDI files rendered to audio were not always identical to those available for download. I sometimes made a cursory effort to tailor them to the rendering synth, making a few adjustments to the initial channel volume balance and instrument selection. The renderings from MIDI to audio were accomplished using a small (5.5 MB) General MIDI patch set, and the MP3s are only 64-kbit mono encodings. Given the sketchy nature of the MIDI input, it would be hard to justify anything more bandwidth-hungry. Page created 27-Apr-1999.  See map for file modification date.