project journal
(November 12) — Two inches in the rain gauge this morning — and this was the view (more or less) as I completed the outline for the project yesterday, in the uncertain shelter of a big Western Red Cedar, looking over a rising Kelly Creek. The rain was nearly over by the time I finished and left, but the sound of the drops on the wet leaves was impressive.
I broke down and ordered a suitable digital recorder today — putting off some other upgrades. So the next time field recordings are called for, I won’t be dependent on Voice Memos and my aging iPhone.
(November 11) — The song remains the same — more or less. All of the various tracks that will make up relays & reconstructions are instances of a single composition assembled in Wotja, the generative music software I’ve been using on most of the hors du troupeau tracks. But because the software introduces gradual and partially random development into the composition as it unfolds, the sameness is relative — particularly as that development is allowed to continue. So, in order to explore the extent and character of that development it has made sense to allow the development to continue hour after hour, but also to record multiple instances of a “single” composition at similar lengths, both as audio files and as midi files.
The Wotja mix, “Positive teal,” consists of a complex lead tracks featuring fairly generic synths and a bit of bass, plus three more tracks of pretty, but not particularly distinctive synthesizer ambience. In this case, I did less tinkering with the choice of instruments than I have in other experiments, turning my attention to the composition rules and fine-tuning the expression of the instruments until I had something I could happily listen to for the hours that would be necessary — as many as twelve and a half so far for a single instance of the mix.
The idea for the relays has been to record four roughly continuous hours of audio for each of four instances, an hour at a time, with a few bars lost here and there in the process of stopping and restarting the recording process. Wotja lacks a pause button, but the very slow development of the pieces means that the lost bars are unlikely to be missed. I’m also trying to record at least one hour from each instance when the mix has had much more time to mutate — after the eighth hour, with less care given to matching up the times. (The logistics get complicated anyway as the process drags on.)
The sixteen hours of “matched” recordings will be split up into twelve-minute segments and then rearranged, so that the first segment of the first instance leads into the second segment of the second, then to the third of the third and the fourth of the fourth, finally cycling back to the fifth segment of the first and completing an hour-long “relay” — or perhaps, to misuse the metaphor a bit, a “leg,” as the four resulting compositions will each consist of four hours, twenty tracks, continuing the pattern until the first relay ends with the twentieth segment of the fourth instance. The second relay will begin with the first segment of the second instance, while the third relay will begin with the first of the third and the fourth with the first of the fourth.
12341 | 23412 | 34123 | 43214
23412 | 34123 | 43214 | 12341
34123 | 43214 | 12341 | 23412
43214 | 12341 | 23412 | 12341
A fifth relay will mix the “late” recordings from each instance, with the lack of “matching” adding some additional randomness to the exercise. My initial thought is that it will take the form of four hour-long mixes. I’m still learning how long Wotja will continue to generate music on various devices — and hope to draw the recording for these mixes from as deep into the process as the software will allow.
The result of all this will be twenty hours of music that is all recognizably connected, developing along lines established by the rules set up for “Positive teal,” but presumably also diverse enough, particularly when the recordings are shuffled together, to periodically underline the developments, chance elements and divergences present. One of the challenges of listening to generative music, particularly of the ambient variety, is that change is fairly constant, but also often infinitesimal, so it is difficult to come away with any clear sense of the musical journey you have been on. That doesn’t seem to me to be a particular problem, but I will admit that listening to really fine examples of long-form music of this type — Marsen Jules’ 24-hour version of “The Endless Change of Colour” comes to mind — the temptation to find other ways of arranging the listening experience has at times been strong. Having now experienced multiple multi-hour instances of “Positive teal,” I’ve reached a similar point — and the relays seem like a potentially useful response.
The reconstructions are similar to those I have done in the past, except that in this case I am making the effort to work with much more faithful midi recordings of “Positive teal.” Again, I’ll make four separate recordings, but each instance will also be reshaped into something rather different from the original, with new instrumentation, effects and mixing done in Garageband.
I’ll be documenting the process with a series of journal entries, spread — according to the undoubtedly idiosyncratic logic driving this project — across the sections for the relevant compositions.
(November 13) — I’m starting to think seriously about the difficulties of syncing up the tracks. There is some slop in the process, thanks to my somewhat limited setup. And, even in the early bars of each instance, there ought to be enough individualization produced by the software to make the initial lining-up of the the instances at least something to be treated with some care. I’ve produced a number of five-minute Wotja-generated audio files, which I’ll carefully compare before starting to arrange the material for the relays.
— With recording glitches slowing down work on the relays, I’ve been focusing quite a bit on the problems posed by the reconstructions. One of the obvious issues is that, while “Positive teal” flows along, in its gentle, indistinct manner, quite nicely over long periods of time, blending into the background as ambient music should, the same pattern of midi instructions, when given different instrumentation and effects, is obviously not always going to provide similar results. Part of the fun of previous reconstructions — particularly “falling back” and “ramblin’ rounds” — has been managing the effect of replacing the gentle synth/bass/horns combo in the main rhythmic track for something more distinct — piano, guitar, etc. — and still working toward something that will, when stretched out to an hour’s length, gradually become ambience (at least for those, like me, with a taste for drones.) With these new tracks, I’ve been using the process as a way of really getting to know the instruments available to me in Garageband — including many that have proven wildly inappropriate for the project, but have still provided some chuckles.
— I’ve got a rough mix together for the third reconstruction — a rather bouncy track built around organ, vibraphone and a synth-string appegiator. Despite the giddiness of some of the elements, I think I’ll be able to give this one an effects-light final mix.
(November 14) — I’m uploading my short comparison files, including some mixed and unmixed versions, once the morning’s recording session is finished. In theory, I’ll have more than three-quarters of that work — sixteen of the twenty hours necessary for the relays — done early this afternoon. And perhaps I’ll make a bit of a push and try to record those last four hours before calling it a day. Otherwise, I’ll probably start chopping out likely sections of the tag-end recordings and start assembling the material for the final relay. I think perhaps also, now that I’ve committed myself to some kind of random assembly method for that part of the project, that I will record more than just the requisite four hours of material — but a lot will depend on just how interesting the results turn out to be.
— I’m also hoping to make a start on the first of a series of lab reports — essentially a revival of the old weekly journal-post format, with some of the “rambles” tossed in — which will summarize the week’s work on various projects, capture the best bits from social media interactions, etc.
(November 15) — A late night’s work produced eleven hours of audio — with results, in general, well above my expectations for the project. I’ve been able to paste the appropriate audio into tracks with the fade-in / fade-out patterns already automated, which has saved some time — although Bandcamp seems inclined to interact badly with that sort of thing at times. I’m keeping the processing simple: individual tracks are untreated; EQ and compression on the master track of each section remains constant; and then each section gets a different reverb treatment with Space Designer. For the first relay, I’m using “large” and “outdoor” settings. For the second relay, the settings are “large” and “indoor.” The spaces modeled are smaller in the third relay and stranger in the fourth. For the final relay — well, who knows at this point what might emerge? I’ll play it by ear.
— I’ve posted full audio for the first reconstruction. Two more reconstructive mixes are close to finished, but I’m still wrestling a bit with what “finished” means in this case. For the fourth reconstruction, my intention has been to produce something considerably more like “Positive teal,” but I’ve wanted to hear enough of the relays in relatively finished form to make sure the final reconstruction still breaks new ground, even while attempting to be a bit more faithful.
(November 16) — I finished assembling the first four relays last night and am spending some time listening to the first relay recordings now. The results are much as I had hoped:
- The transitions from instance to instance are clear, but gentle in the first hour, where they are marked by a fade-out / fade-in transition.
- In the second hour, the differences between instances are sometimes more noticeable, but the slight crossfade is an effective transition.
- As the tracks progress, there is an increased divergence among the instances, but the combination of generative progression and applied reverb also makes what was initially at least a kind of tune into something more like a general tendency, expressed in an increasingly muted and random manner.
- By the third hour, the more melodic elements from the rhythmic track tend to come and go, sometimes with new melodies emerging and sometimes with familiar motifs reappearing — while the more ambient tracks follow a general rule of “mutate, but-harmonize.”
- And, while there is a lot of development between the fourth hour and the eleventh, by the fourth things have, for the most part, settled down into a kind of rich sonic soup.
(November 17) — With final relay seeming lovely, but a bit too wallpapery, I opted to craft a version of “Positive teal” (“Today painting,” according to Wotja) suitable to supplement the tag-end recordings in a new relay. I’ll get those rough mixes saved today, before returning to the “Constructing Anarchisms” glossary project. But I also let the new mix be my sleeping music for the night and am now listening to the ninth hour of it. This time, I explicitly planned for it to be pleasant and interesting as a full mix, but also sufficient for separate listening in pieces, with individual tracks having a comparative quiet, sparse character compatible with those rather washed-out tag-end recordings. There is still a lot to learn about the technical aspects of using Wotja, but this experiment was more about understanding the general dynamics of a particular generative development — and it feels like a success by that measure.
(November 20) — All of the projected tracks are now complete and posted here, along with quite a few extra bits. I’ll probably return to this page after I’ve had a chance to live with the tracks for a while, but, for now, that’s a complete project.
first relay
12341 | 23412 | 34123 | 41234 — 4 hours
(November 11) — The First Instance was an audio recording (on my desktop) of audio from a first-generation iPad Air. It seems to be able to generate about twelve and a half hours of the mix before the audio shuts off, so I have been able to listen to instances of the mix as background music through daily tasks or as background noise at night for sleeping. I made the initial four hours of recording with minimal incident — and recorded an additional hour at the eight-hour mark and another between 11:17 and 12:25.
(November 13) — Argh. Yesterday was ruled by the digital gremlins, at least where the audio recording was concerned. I had to replace the first First Instance recordings with some made via the desktop — and now have reorganized and backed up my files in a way that should be relatively gremlin-proof.
(November 14) — The audio tracks are all recorded and I’ve started to mix some of the sections. Just because of the complexity of the project, I may end up mixing the first hour of each relay before moving on to the second hour, etc. And the mixes will vary. For the first hour of each mix, the switch from one instance to the next will be pretty precisely at twelve-minute intervals, with no cross-fading and just a short fade-out / fade-in to mark the transition. For each subsequent hour, the transition will include increasing amounts of cross-fading and will at least potentially be shifted a bit off the precise interval in order to smooth transitions — minimally in the second hour and perhaps quite considerably, in all those respects, by the time I get to the final relays. I’m using Space Designer to shape the general acoustics of the segments, with the first relay getting various sorts of outdoor treatments.
— Listening through the first “leg” of the first relay, I’m pleasantly surprised by the overall effect. In hour upon hour (upon hour…) of listening, I’ve come to think of “Positive teal” as pleasant, but perhaps a bit innocuous. It hasn’t taken more than just a bit of gentle processing to make it rumble and occasionally roar.
(November 16) — I’m posting first relay, after a morning listen-through. This time around, I’m particularly struck by the end of the third hour, which moves from quiet and beautiful to a bit ominous, and the beginning of the fourth, which is very quiet. As a whole, this first set of tracks feels more like an album than I had initially dared to hope. I’m remembering the process with the Guinea-Pig Fleet tracks, back in the day, when, after a lot of up-front planning, the success or failure of the whole was often dependent on more or less aleatory process — so that there was a kind of first discovery of the result of processes that you had put into motion. This is very much like that. And while I am still a bit rusty with some audio skills, I’ve had a long time to thing about patterns, chance, ambience, etc. — particularly through this strange pandemic period — so, in some ways, maybe I’ll a lot better prepared than I was then. In any event, despite considerable differences in the sonic palette, this feels very much like the Guinea-Pig fleet recording that I never quite made back then.
— One last tweak, after the initial upload. The ending was a bit abrupt, so I’ve made use of a couple of minutes of audio available at the end of two of the instances and, I think, done a bit more justice to the finish.
second relay
23412 | 34123 | 41234 | 12341 — 4 hours
(November 12) — I’m spreading the work over different devices, mainly just because they are available. A played an instance of “Positive teal” on my third-generation iPad as sleep music overnight and am recording the final (?) hour as documentation. It’s nice that Wotja will run in the background on the more recent tablets, which isn’t the case with my oldest Air. When this instance ends I’ll restart things and record the first four hours of audio for the Second Instance.
— Just under twelve and a half hours does seem to be about the limits of how long Wotja can extend an instance.
— The digital gremlins made off with an hour of recorded audio, so I’ll restart this instance tomorrow. But the interruption prompted some rethinking about logistics and the process for the final relay.
(November 14) — As I start to mix the tracks for this second relay, I’ll be using a selection of indoor setting precepts in Space Designer — nothing cavernous, but the tracks will certainly have some opportunity to breathe freely.
(November 16) — The second instance of “Positive teal” starts in a place more like the middle of some of the other instances, which is nice in the context of the larger work here, since, while many aspects of the development have been taken back to zero, there is a still a sense that we haven’t returned to the beginning of first relay.
— It really isn’t until “two hour four” that the motifs — particularly prominent in the first instance and, as a result, in “one hour one” — return in a dominant way, bringing things back to more distinctly familiar ground. I like the more meandering music in the first three hours of this relay very much, but I might have considered it something of a failed experiment if that return had never occurred.
(November 17) — second relay is now available online.
third relay
34123 | 41234 | 12341 | 23412 — 4 hours
(November 14) — The Space Designer settings for this relay model somewhat more confined spaces — although the basic tracks already have a rather airy feel much of the time.
(November 14) — Having recorded the various tracks and decided on the rule for organizing the relays, the toughest decision was choosing an order for the instances. In theory, it was just going to be a matter of the order in which the tracks were recorded, but complications with the recording process itself necessitated some changes. One of my files ended up with the name “Instance five or one.” I knew that whatever ended being the first instance for the purposes of the relays should at least have a beginning that sounded a bit like a beginning — not like the beginning here, in what became the third instance, which is great in large part because it doesn’t sound quite like a beginning.
Things seem to have worked out pretty well on that front. And my greatest concern at the start — that perhaps there just wouldn’t be enough variation in the instances or enough development through the four hours — was unnecessary.
(November 19) — On and on and on… The difficulty at this stage is maintaining some relatively clear memory of all that has come before as I listen to the successive recordings. At some point, that’s not really a realistic expectation — and reaching that point in a way that seems natural will be a certain kind of success. Midway through “three hour four,” I’m feeling like I’m about there — fairly successfully.
fourth relay
41234 | 12341 | 23412 | 34123 — 4 hours
(November 14) — I’m going to try to remain tasteful with the Space Designer modeling here, but in the sense of sticking to the least extreme of the “warped outdoor settings.” The final relay probably won’t call for much of that kind of processing anyway, as the generative rules will have been doing their work for eleven+ hours on all of the source material, so this is probably the spot to let these earlier recordings explore some of the (even) more unconventional possibilities.
(November 19) — This is the point in the process where it’s possible to breathe a bit, look back and say: Huh. This unlikely experiment seems to have played out pretty well. The sixteen hours of recordings, methodically shuffled together in 12-minute segments, work pretty well, whether you’re considering them as eighty individual songs, sixteen hour-long albums, four longer compositions or one audio monstrosity — itself just a part of a larger composition. A lot of that is obviously a function of the software’s capabilities — three cheers for Wotja and the folks at Intermorphic! — but at this point I think I can pat myself on the back, at least in passing, for having made some good guesses in the earliest stages of this affair.
— I have at least draft versions of all of the tracks loaded into Music now and a playlist of the whole mess assembled.
“33 SONGS — 1.2 DAYS” — I couldn’t suppress a real laugh when I first saw that.
— A happy accident: The last-hour recordings of several instances run a bit long, so it has been necessary to trim some of them back down toward the hour mark in various tracks. In the case of “four hour four,” there were two minutes of the first instance that might have been trimmed, but it turns out that they mark a return to the more familiar motifs, after a final segment that was particularly indistinct, so I worked them into the mix, as a kind of return to something like the beginnings of things.
final relay
tag-end recordings + new version of “Positive teal” — 4 hours
(November 12) — I’ve opted to use the very tag ends of the audio Wotja is capable of producing, starting around 11:15 and recording until the music stops. But lots of things can interrupt a process of that length, so my goal is simply to get four good hours of end-of-the-line audio, whether or not it comes from the same instances used in the first four relays. That late in the process, the instances have presumably diverged as much as they are going to, but each instance also shares, in those final stages, a soft, indistinct quality. So perhaps the right game to play with this final relay is to recombine the instances in ways that test whether than shared softness reflects other shared characteristics. I’ll play around with methods of sequencing sections of these recordings, but I’m inclined to think that perhaps just allowing the Music app to randomly play and crossfade the sections might be as satisfying an experiment as any.
(November 16) — Working on the segments to be mixed from the tag-end recordings, it’s clear that, while there is considerable diversity across the recordings, the indistinct quality of the material may make more than one or perhaps two hour-long mixes of the material excessive. So I’ll probably do one untreated mix and then treat the other three to some more interesting forms of processing.
— On second thought — or third or fourth at this point — I’ve decided to produce three extra hour-long instances of “Positive teal,” isolating the various parts of the original Wotja mix and mixing up the instrumentation.
(November 18) — These tracks are now roughed out. I tried to walk a fine line between maintaining the character of the original mix in the new version of “Positive teal” and adapting it to complement the tag-end recordings from the first version. First, I eliminated some elements that hadn’t produced much audio anyway. Then, I spent some time muting and soloing tracks and elements within tracks, until I was pretty sure that I knew which were producing the most distinctive motifs in the first version. With those elements intact and dominant in the new mix, I just moved things around and added minor bits until I was sure that I could break the mix into three separate parts, each to be recorded separately, without ended up with too much cacophony or silence.
— There are indeed a few moments of silence.
(November 19) — The first section of “tag-ends and new beginnings” gets off to a bit of a slow start. This is what eleven hours of development will do to the mix. But it manages to provide muted versions of a number of familiar elements, before transitioning into the first of the new tracks. And that transition is rather nice. I’ve listened to days and days of the late stages of “Positive teal” and continue to find it pleasant background noise, but there is always the sense that what I’m listening to is indeed the ending of something now only recognizable in the general tone of things and the more distinctive bits that sometimes rise to the surface. Using it as the basis for building something new, gradually adding in the pieces of the new arrangement of “Positive teal,” is fun. And, of course, that new arrangement never appears as in undivided form here. I was undoubtedly add at least one recording to the supplementary tracks compilation that seems to be forming. But, as these tracks now seem to be a kind of bridge between the relays and the reconstructions, perhaps that whole version would naturally follow this somewhat piecemeal introduction of its elements, with or without some number of intermediate constructions.
— I’m finding these recordings fascinating to listen to, having just made my way through the first four relays in a relatively short period of time. After listening to the earlier recordings for a while, you get sensitive to the possibilities of which elements will come to the front of the mix next. It’s hard not to anticipate in much the same way here, but because of the way that the elements are divided, the expectations are frequently disappointed. Instead of the regular interplay between elements, that sort of careful attention is likely to instead find itself focused on the continued development of those few elements at play in any given section.
first reconstruction
1 hour
(November 11) — This began with a midi recording made on my desktop computer. A careful export provided me with 14 tracks, three of which didn’t contain enough to worry about. The obvious “lead” track was a bit sparse, but, once I had decided to use an acoustic guitar for that, I was able to find a compatible arpeggiator and a track that allowed it to fill things in a bit. Another arpeggiator gave me a rhythm track and I opted for a variety of keyboards — heavy on arp patches this time — and a harp to flesh things out. The resulting ensemble is perhaps a bit unlikely, but I’m finding that the upbeat mix of sounds that came together is wearing well through repeated play. I may wait to have more of the pieces together before I really mix this track, but it’s nice to have at least one piece of this project at least complete in rough form.
(November 19) — I have experimented with quite a number of different ways of approaching these reconstructions. With earlier tracks of a similar sort — “falling back” and “ramblin’ rounds” — listeners had no experience of the track in its original form, except as it appears mixed in with the new instrumentation. But in this case, a listener might have already experienced twenty hours of material derived from the mix to be reconstructed. It seems important to maintain some connections to the source material, even though, in this case, it won’t be part of the mix.
(November 20) — I ended up with two reconstructions, the first and third, which were both interesting, but in very similar ways. I’ve included both of them here and will start from scratch on the remaining track, perhaps beginning with midi tracks from the second version of “Positive teal.”
second reconstruction
1 hour
(November 11) — The second set of midi recordings on my desktop are surprisingly sparse. I’ve started a third for comparison purposes, but it looks like sometimes the differences between instances are really dramatic at that level.
(November 12) — I’ll have the new set of midi recordings ready to toss into Garageband by the time my current audio recording is done. Then maybe I’ll have a first look before recording the Second Instance audio. And maybe, before things progress much farther, I’ll just recommission my old Mac Mini and split the work. The promise of Universal Control has already set my thinking in that direction — and I already have the Logitech equivalent set up, so it’s just a question of reassigning monitors.
— The third midi recording looks much more like I expected it to, so I think I’ll assume operator error for the time being and work with this material for the second reconstruction.
— And it looks like this reconstruction may start as a woodwind quartet.
third reconstruction
1 hour
(November 12) — Midi tracks are recorded.
(November 13) — The differences in the midi recordings are obvious as I try to give them new instrumentation. I’m doing my best not to simply make things “work” by drowning everything in reverb.
— Having spent some time comparing multiple recordings of the first five minutes of “Positive teal,” it’s clear that some instances are a lot sparser than others. And, of course, the way that I am currently recording midi tracks means that I am, in some sense, already combining different instances of the same Wotja tracks. At some point, I need to set things up so that Wotja is driving track-generation in Garageband directly, but that part of the plugin interface is so far a bit beyond me. Anyway, the tracks in hand for this fourth reconstruction are perhaps not so promising a “seed” for reconstruction, but I’m inclined to try to solve the problem by getting more sophisticated in my attempts on this end, rather than recording any new tracks.
(November 15) — This track, which was intended to be something of a return to the original “Positive teal” sound, has been driving me a bit crazy. Wotja is particularly good for constructing complex, layered ambience — while the selection of instruments already loaded into my Garageband (my palette for these tracks) excels, let’s say, in other things. But it turns out that the key to begin taming this track was finding a compatible piano.
(November 20) — This track, which started out as the fourth of the reconstructions, has become the third, as the original third becomes a bonus version of the first.
fourth reconstruction
1 hour
(November 12) — Midi tracks are recorded.
(November 13) — With additional complications emerging in the recording for the relays, I’ve moved forward with the reconstructions, establishing rough instrumentation for this third track and getting a start on the fourth.
(November 13) — While I produced a third reconstruction that I like, it is similar enough to the first that I have posted it above, with that track, as a bonus and will start again fresh.
(November 20) — Back to the drawing board it is, using midi recordings of the second version of “Positive teal.” And it looks like this mix will play with all-synth instrumentation, heavy on pads, preserving the general rise and fall of the various tracks, but not a great deal more of the original dynamics.
— One of the things about Wotja compositions is that the software is quite good at making lots of elements harmonize, so when you start to work with the midi information there can be quite a lot of it. The challenge — and much of the fun — in these reconstructions has been transforming what are potentially very busy mixes into tracks that demonstrate the complexity of the source material, but don’t just devolve into chaos. That challenge isn’t diminished at all when the instrumentation is going to be mostly synth pads and soundscape patches. It is still necessary to keep the sounds varied. So this track took some careful handling in order to avoid the muddy wall-of-synth that it was always threatening to become. Ultimately, the result doesn’t put a smile on my face in the way that the first reconstruction does, but it does strike me as another successful use of the source material that demonstrates the qualities of “Positive teal” in yet another way.
Just for reference, here are some additional, untreated one-hour recordings of the two different versions of “Positive teal.”
First version:
Second version:
