Pure Data algorithmic composition in 53-EDO (with visualization)

[@mattrist]

I'm 8 years late to the party but this is incredible...

[@Sevish]

Love your work so much!

[@scottjampa6374]

sevish hi!

[@jacobkantor3886]

Sevish i love you work so much

[@acreil]

I'm working toward more exotic sounding stuff; first I wanted to see if people would find this palatable.

[@polifemo3967]

this is more than palatable, I actually like it!

[@adilsarbay3181]

@@polifemo3967 Agreed!

[@SCWood]

I love this. You should make an album of this ambient microtonal stuff.

[@acreil]

@@SCWood I usually make stuff too noisy to count as ambient.

[@hernanescudero6620]

Dude put a 4/4beat to this and millenials will dance this

[@muteqx]

The chord that occurs from 4:38 to 4:59 is really delicious, and then the little meandering melody from around 5:43 is pure magic!

[@carmushroom]

"microtonal algorithmic composition" sounds like something that would be cool in concept and unlistenable in practice, but this is super good. very organic and full sounding.

[@henrikljungstrand2036]

Yes it is almost like listening to natural sounds, you know ocean waves breaking, leaves rustling in the breeze, rain pouring down, birds chirping etc. I suppose the algorithms to some extent emulate natural processes.

[@Blue-Maned_Hawk]

It sounds like the type of thing that would be technically impressive but not enjoyable, like downhill skiing or certain kinds of jazz, but yeah, no, i agree with you. This is good shit.

[@BenStoneking]

You're ability to combine mathematics and music and create something truly beautiful is astounding. You don't come by people like you everyday. Most pure data nerds just make crazy weird noises. But I could actually listen to this any time.

[@acreil]

I make crazy weird noises too though.

[@Gesceap]

You mention: "Most of the sounds are made by my own algorithmic additive/formant synthesis engine. I might do a video later that describes that." I would love to see that.

[@acreil]

I stopped using it because it was unwieldy and sort of stupidly implemented. It looked cool and did some neat things, but everything used messages, which wasn't really a good idea. I still use some things derived from it, but it has to be repatched every time I want to change something, so I think it's of limited educational value.

[@Gesceap]

acreil I would be very interested in seeing anything related to your workflow and how you use puredata.

[@acreil]

I concluded that the only way to communicate anything useful would be to live stream the creation of a patch in its entirety. I'll try to do that eventually, but I'm busy recording other stuff now.

[@Gesceap]

acreil understandable. I look forward to watching it as well as your other music.

[@rubensesaki663]

Man, I don't know how I ended up here. I don't know much about microtonal stuff, but this song helped me go through an anxiety episode. I listened to it twice right now and I can't describe what I fell. This is very beautiful, I don't know why

[@dream.machine]

Come back and listen to this in the 2040s - 2050s

[@HW-ct1iq]

This is making me download and (attempt) to learn to use Pure Data. Thank you. This is beautiful.

[@SimpleTrax]

If timed right, playing two of this video simultanously creates nice effect as well.

[@microtonalmilio5233]

When you started one video, how many seconds do you wait before pressing play in the 2nd video to achieve this effect?

[@SimpleTrax]

@@microtonalmilio5233 Seconds not so important, but try to sync the metronome "tsh tsh tsh tsh", metronome seems to shift slightly also..

[@rrr00bb1]

I like the sound. 53ET is most usefully thought of as a cutting of the Just whole tone (9/8 ratio) into 9 equal parts. If you build scales by walking out 22 "frets" (ie: by fourths..very close to 4/3), then a span of 7, gets close to a Just diatonic scale. 6 of these whole tones (ie: 9 * 6 of them) overshoot the octave by 1 interval. 53ET is actually used in the real-world; Turkish master scales, though usually as an almost 12-et fretting where there is an upper/lower half tone choice.

[@nsrocker99]

Just started messing with pure data. This is such an inspiration!! Thanks for sharing.

[@acreil]

No. 53-EDO does a good job of sounding "normal", as it's a very good 5-limit temperament. All the intervals that you're used to hearing are well represented (shown as 5, 9, 14, 17, 22, 26, 31...), and these tend to be favored as they're mostly the least dissonant ones. But a lot of intervals do appear that aren't available in 12-EDO (10, 12, 19, 24...). The result is fairly conservative sounding; I didn't want it to be too jarring. But in general most of the scales don't have 12-EDO equivalents.

[@yourtypicalcupoftea]

53edo’s subset of notes called Orwell is great at making grieving and bittersweet sounds

[@DigitalBuoy]

That has to be the nicest thing a stranger has said to me about my music! Thanks so much!

[@acreil]

Right. The intent is to have some unique or novel qualities without it being overly weird. Some of my earlier attempts at that have been rather less successful, even in much simpler temperaments (see the 19-EDO one). That's not to say that I want everything to be ordinary sounding, but I'd like to use each temperament for what it does well. If I want dissonance or tone clusters, I can use random frequencies or some arbitrary invented temperament.

[@SCWood]

This is fantastic. It sounds so dissonant, yet so relaxing

[@guerrillaradio9953]

Pure meditative bliss. Thank you.

[@hizzogg]

Wow! I love it! I could listen to this for days.

[@JohnSmith-iu3jg]

In all honesty though, 53ET is very very accurate to both 3-limit and 5-limit tuning systems. It also allows good intonation for other important intervals such as the 7:4 interval, also called a septimal (7-limit), or harmonic, minor seventh (B7b).

[@lychee5269]

John Smith What is the best limit tuning??? Like what is the best number to use?

[@AttitudeCastle]

@@lychee5269 They are all different with different feels and sounds, just choose on you dig and run with it!

[@henrikljungstrand2036]

@@lychee5269 There is no single best tuning. And not all tunings are either equal temperaments or just intonation, there are lots of useful linear temperaments and quite a few planar temperaments etc. Every regular temperament is defined by a Just Intonation prime limit (or seldom a subgroup limit), plus a set of "commas" (small intervals) tempered to union, usually this set is generated from an explicit set of generating commas. In 2-limit you only have octaves; in 3-limit everything is basically about octaves, fifths, fourths and Pythagorean major seconds. In 5-limit pure major and minor thirds and sixths enter the scene, plus minor sevenths, smaller major seconds and major sevenths. In 7-limit we also get pure subminor and supermajor seconds, thirds, sixths and sevenths, plus pure lesser and greater tritones. In 11-limit we get semi-pure neutral seconds, thirds and sevenths plus superfourths and new minor sixths. And in 13-limit we get semi-pure neutral-ish seconds, sixths and sevenths plus subfifths, subfourths and ambivalent second-thirds. Not much of interest in higher prime limits, except tame dissonances. In the 13-limit i think the best temperament is 270 equal divisions of the octave, it is accurate enough to sound almost completely like just intonation, yet pretty easy to tune since it is an equal temperament, also it doesn't have EXTREMELY many notes (like 2460 edo or 8539 edo), plus the step size is about the size of the just noticeable difference of the average human ear/brain, also it is consistent in the 15-odd (16-integer) limit (and probably higher than that), and last but not least it tempers out extremely many small superparticular 13-limit commas, while not tempering out the large ones (676/675 being the largest superparticular comma tempered by 270 edo, much smaller than 81/80, 126/125, 225/224, 99/98, etc of Meantone such as 31 edo or some other tuning of it), making it support lots of accurate linear temperaments like Ennealimmal (or Hemiennealimmal and Semihemiennealimmal), Decitonic etcetera. Most people still think 270 edo is too many notes, so they settle for say the excellent 72 edo or 87 edo or 68 edo, or something smaller like 53 edo or 41 edo or 31 edo. Some people like less accurate temperaments that are still more expressive than 12 edo, like e.g. 28 edo, 27 edo, 26 edo, 22 edo, 19 edo, or 15 edo. Or even the really flat 23 edo or really sharp 18 edo. Some people use linear temperaments with less notes, like unequal Semantic Daniélou-53 (Pontiac) of 53 out of 171, unequal Magic of 22 out of 41, unequal Meantone of 19 out of 31, or unequal Pajara of 10 out of 22, or even planar temperaments with even less notes yet greater accuracy like very unequal Marvel 12, 19 or 22 out of 197.

[@henrikljungstrand2036]

I still think 41ET is better for accurate 7-limit with manageable number of notes (72ET is more accurate, but has many more notes). 65ET is arguably better in 5-limit, and 118ET way better, followed by 171ET in both 5-limit and 7-limit. 53ET is obviously really good in 3-limit, this limit becoming better only in 306ET (or double that for 612ET in 5-limit or 7-limit), 359ET and especially 665ET which is practically pure Pythagorean tuning (still pure after cycling through 665 fifths, getting back to something completely indistinguishable from a unison).

[@zhou_sei]

beauty

[@DigitalBuoy]

Lol, yeah i remember seeing your "How to make music with 2 euro-cents and a lighter" vid, interesting stuff

[@acreil]

Each piece is mostly made from scratch; I don't reuse much as I generally want to try something different each time. I do some editing and mixing, but mostly it's just the software doing its own thing. Anyway the question of authorship and intent isn't really something I'm trying to address. As far as I'm concerned, my intent is expressed in the sense that I messed with it until I was satisfied. Of course that gets muddled if others start using and modifying the algorithms...

[@петрзадунайский]

the music never ends

[@Ivan_1791]

That sound are so nice.

[@cbmtrx]

My question about algorithmic music in general is: how much musical intent is left in a work after the composer has completed the code? Is the output strictly code-based or are you involved, creatively, with the whole form of every piece that is yielded by the code? Pretty fascinating, either way.

[@cbmtrx]

Actually I think the phrase "sensible results" in your annotations is where I'm finding a conflict between the intention to expand the range with 53-EDO while ultimately reducing it to have it sound tonally...acceptable? consonant? meaningful? to western ears. I would imagine that outputting a full, unmolested 53-EDO tonality would be...very dissonant and a challenge to make sense of!

[@geecen]

Very good. I'd definitely to have a look at the synth engine too.

[@cbmtrx]

Interesting. I think to a western ear this sounds tonally "consonant" with the "dissonant" (non-12-EDO-like) bits merely being perceived as acceptable vibrato and/or glissando artifacts that would naturally occur in 12-EDO tonality during performance. I think the "humanization" aspect of normal performance keeps this well founded in a 12-tone feel. That being said, I love your formant synthesis, which gives this a gamelan-like mood. Reminds me a bit of the soundtrack to the remake of "Solaris".

[@acreil]

I've posted abstractions and various portions on the PD forum, but for the most part I think the full patches are too messy. I'd like to post more, but generally I end up just moving on to the next thing I'm working on.

[@wjoojoo]

Love this!

[@ralfyrules]

this is awesome, Great work!

[@JTylerBentley]

Also, those annotations were very helpful.

[@carltaylor43]

very nice! Im currently researching generative techniques... I might take some influence from your method here :)

[@RutgerMullerMusic]

Beautiful!

[@im_alivemusic2161]

very inspiring, thank you!

[@blueschase11]

I think all music should be in 53 edo it's so fantastic.

[@scottjampa8308]

you can still get to turkey with a US passport lol

[@henrikljungstrand2036]

53 is fantastic but 171 edo is better. Almost pure 7-limit just intonation. For almost pure 11-limit go to 342 edo (double the steps). For 13-limit 270 edo is excellent, slightly less pure than 342 in lower limits, but much better than it in this limit. I suppose 118 edo is as good as 171 edo for 5-limit. There is a chain of 53 fifths in 171 edo (and 118 edo, and obviously in 53 edo) that yields 8/5 after 8 fifths (reduced by 4 octaves), and 7/4 after 39 fifths (reduced by 22 octaves). Thus 8 fourths yield 5/4 and 39 fourths yield 8/7 (assuming octave equivalence). This is described in the Semantic Daniélou-53 scale, which i think will be of great interest to you: www.semantic-danielou.com/semantic-danielou-53/

[@polypx]

I think this is really beautiful. There's a lot of consonance outside 12et to be discovered as well as dissonance. (There's a lot more consonance outside 12et than within it, if truth be told.) Anyway, brilliant. I'm going to chase down some more of your pieces.

[@florishenning5089]

WTF! This is amazing!

[@JTylerBentley]

This is very cool.

[@JuliusRaskevicius]

impressive.

[@rrr00bb1]

19 isn't really a simpler temperment. It has fewer notes to choose from, 19et notes are too far from where they are supposed to be (according to Just Intonation). 53et is so close to 5-limit, that in a way it is actually simpler than 12-et. It lines up so well with the physical reality of sound that the justifications for 12et seem convoluted by comparison.

[@thesoftdistortion]

This is lovely. :-)

[@Ykulvaarlck]

considering youtube purged all annotations, could you provide an explanation for the visuals? it'd be a shame for that to be lost

[@acreil]

I don't remember much specifically, but basically the scale is numerically represented by the squares on the diagonal, and the other squares show the intervals between those pitches. Colors are derived from the interval size, and the highlighted squares show the intervals in the current chord. Ultimately I didn't think the idea was all that useful, since there's not really an obvious correspondence between what's seen and what's heard.

[@DigitalBuoy]

Your brain must be one of the newer upgraded models

[@skriptico]

I

[@JohnSmith-iu3jg]

Oh Jing Fang, what have you started

[@veronikatabakova8755]

Amazing! :) Love it!!! Could you share some experience? What did you use in Pure Data? :)

[@billystiltner]

nice

[@cbmtrx]

Lovely sound here; an intriguing quality. But I have a question: for what is described as "53-EDO" this sounds just plainly 12-EDO...or am I misunderstanding the aim? Is the elimination of dissonant scales just yielding a basically 12-tone sound?

[@cbmtrx]

Well, like it or not, you're composing--even John Cage and his chance music would agree with this. I guess my curiosity on the topic is focused on how to produce algorithmic music without giving up important compositional controls... It's a different compositional proposition I suppose.

[@acreil]

working on it

[@erikmaes3427]

Cool stuff. How did you animate this? Is it Gem or clever canvas manipulation?

[@acreil]

Canvas manipulation. It wasn't so clever though.

[@marvotron]

that's beautiful. you could add the PD patch to download, that would help me understand it ;)

[@sachs6]

How do I see said video annotations?

[@acreil]

they're long gone

[@acreil]

I think it's one of the recalled models, actually.

[@romeolz]

"The visualization part is described in the video's annotations." About that...

[@acreil]

It's been gone since forever, I should probably edit that.

[@josephzizys]

nice! do you post your patches anywhere?

[@DigitalBuoy]

Thanks! Have we met before?

[@switchingworlds8751]

paul lansky tier. love it

[@JohnSmith-iu3jg]

Nothing like reducing that ratty Pythagorean Comma to something resonable. "(2)^31=(3/2)^53" ftw

[@GRAVYsailor]

very cool! saw you in the soundcloud thread i'd love to see this patch if you don't mind sharing we're into a lot of the same shit--i've been messing around in pd and i'm also really interested in doing stuff with alternative tunings and scales. plus i see you've got an ht-6000! i'm getting an ht-3000 pretty soon i'm over at soundcloud. com/tenzingnorgay btw, don't have much music up now but that should change sometime soon

[@hypnovia]

How do you calculate "total" dissonance?

[@acreil]

The idea is to first calculate, for each interval, the complexity of the nearest just interval (this is the product of the numerator and denominator) and error (the deviation from that interval). I mess around with that until the results more or less correspond to my subjective impressions of what's more dissonant vs. less dissonant. Then for each scale get the total dissonance by summing the dissonance for each interval in the scale's interval vector. But since I'm treating all notes of the scale as equal and not making any assumptions about the scale on the PD end, I think it works better to use the sum of the square of the errors. That way I don't end up with scales that are mostly good but have one really horrible interval. Then I sort the scales in order of increasing dissonance. If the temperament has passable thirds and fifths, major scale is the least dissonant.

[@hypnovia]

How many just intervals do you use for reference? I feel like the more you use, the better - but isn't there some point at which it isn't worth computing any more?

[@acreil]

That part is kind of tricky. I can set some kind of maximum complexity ceiling based on prime limits and so on. But what's the highest prime limit that's perceptually meaningful? Maybe 23? Or 31? Another problem that comes up is that for higher EDOs, multiple tempered intervals are going to map to the same just interval. So in 1200 EDO (not that this is a useful temperament in particular), 702 cents is closest to 3/2. But 701 cents is also closest to 3/2. It doesn't make sense to treat 701 cents as a distinct interval. So at that point it's really kind of trying to temper the scale. But this presents difficulties when modulating to a different scale. I could imagine that it would be useful to allow a little window, within which the pitch can drift so that all intervals are nicely tempered. But actually trying to implement that would be very difficult. I think it's kind of fudged in some way no matter what I do. I just try to come up with some heuristic that kind of works. A lot of unexpected stuff starts coming up once you try to get away from 12 equal.

[@hypnovia]

What's interesting is that 7/5 gives an interval very near the tritone, which is dissonant. If I were programming this, I would account for the factors of the numerator and denominator; say, 7/5->7+5=12; 8/5->2+5=7; 5/3->5+3=8; 3/2->2+3=5; 17/16->17+2=19.

[@acreil]

Yeah, I think I did something like that. But the product of the numerator and denominator is important too. 729/512 is obviously more complex than 3/2, even though they're both 3 limit intervals. I fudged it somehow, I don't really remember.