Automatic music : is it possible ?

Chords, scales, harmony, melody, etc.
User avatar
Tj Shredder
5057 posts since 6 Jan, 2017 from Outer Space

Post Mon Jul 27, 2020 5:13 am

I can point to some algorithmic music I programmed myself in the 80s. I think there is a danger in focussing too much on AI which tries to mimic existing music. Its way more interesting to invent your own rules...
This piece was completely automatic, the others in that album used the same algorithms, but the form was composed...

13 posts since 9 Jul, 2020

Post Wed Oct 21, 2020 12:36 am

Hello to all,
Some time ago I had opened this trhead on the production of automatic music:

For some time now I have been studying the question of transformations in music. I would like to know your opinion on these transformations, knowing that as always the sounds are not realistic because I necessarily go through a computer.

In the first transformation I radically modify the harmonic structure of the first movement of Beethoven's fifth symphony:

In the second transformation the transformation essentially modifies a major key and a minor key. This is Tchaikovsky's violin concert:

There are other transformations in my youtube channel, which I'll let you discover if you feel like it.

The question is then the following: do you think these transformations manage to produce a coherent musical universe, even if very different from the original musical work?

10 posts since 19 Oct, 2020

Post Wed Oct 21, 2020 2:13 am

I prefer the originals, probably because of bias and familiarity. Though I think the transformations do produce "coherent musical universes". Like every musical piece, they also create their own universes. Some pieces are just more successful and joyful to explore than others. In this case, the Beethoven transformation gives a creepy vibe, almost like an uncanny valley for music. It plays on expectation also; you know the melody, but something is off. Basically the building block for most horror. :)
The transformation appears to be less harmoniously cohesive than the original, and thus less satisfying to listen to, at least to my tastes.

112 posts since 19 Jul, 2014

Post Thu Oct 22, 2020 8:15 pm

Thank you for your work. Tonight I haven't yet listened to your examples; I just read a few pages; but I wanted to throw my two cents out while I am focused on it. I get that certain tones are essential and that others may be transformed. I find that the mathematical connections such as this are real but quite sophisticated - and beyond most of us theorists. But for me it goes in a slightly different direction. For example, if we stick with 12-ET and simple timbres, using such a method as yours, we could arrive possibly at something manageable as far as an AI, but sticking only with 12-ET can cause us to miss some very interesting and extremely musical sounds.

Jimi Hendrix used a 7#9 to great effect, which works particularly well with the harmonics of certain kinds of distortion. Led Zeppelin has a song Dancing Days in which the repeated riff has a very nuanced guitar string bend. The rock genre has so many of these melodic and harmonic progressions, that would not particularly work on piano. It seems to me they arise from subtle shift of melodic scaling (bends and tuning, and vibrato) and timbral variance. Hendrix and Van Halen were masters of bend and timing. Listen to House Burning Down played by Randy Hansen (Hendrix cover artist). Randy did not just need to master the notes and sound fx of Hendrix, he needed to (and succeeded at) mastering Hendrix's ability to slide back and forth in the expected beat flow to give an emotional effect. And this timing can be part of the equation too - things that work incredibly well if timed the way Hendrix did, may not work as well without that "feel". The "transformed" tones may effect this feel more than one expects. I tend to think that these influences sometimes evolve specific pendulums of harmonic gravity, which were given rise to for these composers, that would be lost when played using a basic MIDI piano sound. But I am disappointed that no software is being created that agrees with my principles.

For example Modartt Pianoteq does allow changing the tunings, but it is not strictly a matter of "tuning", since tuning ordinarily is conceived of as static. I believe in a tuning bed that can be varied relative to positions in the musical rhythm, and even to other instruments and with harmonic and rhythmic "applique" (meaning some near-ratios and near-repetitions are more stable and others are less stable - the less stable material is like sewing an additional ornamental piece of cloth on the main body of fabric - and I tend to think this is like what we're hearing in music - the vibrato and chord progressions and rhythms and timbres can be quite rote and basic - but these need not be - and different genres escape obviousness differently by varying different perceptual elements, be they lyrics, timbre, syncopation, time sigs, or thick jazz harmonies ) I favor Helmholz, Hindemith, and Cope and such, in that I tend to believe in harmonic rooting (aka interval roots - not to be confused with the bottom of chords built in stacked thirds - see Wikipedia), resulting in a system of harmonic gravitic forces that could be somewhat predictable (purely in ratio root calculations), until you add in harmonic timbral variation and pitch bend. A pitch I bend to quite often on the guitar is the 7:4 interval, over a Major or minor triad. When bent into and out of this interval sounds "nasty" in a positive way, needing resolution but not so much in a straightforward way as in a way that suggests whole journeys that may or may not arrive Home - harmonically (and if you knew that Home was assured, would it feel this way?)

A lot of guys learning guitar are not always aware that the frets on the guitar are not ALWAYS where the harmonically pleasing sound is. This is because our more common scales result from sequences of Perfect Fifths. And if you think of them that way you can just use 12-ET. What changes this for these developing student musicians (if they didn't start out using their ears as much as they should - thinking too much of theory and patterns) is that they begin to hear the different kinds of "capacitance" that happens in music, along with the inductance of the basic Perfect Fifth tension (which is relative to the very pure sound of 3:2). Then they begin to bend and use vibrato more towards the actual harmonies of vibration. (Harmony could be said to originate from "means of joining" - but in music it is a means of orbiting, de-orbiting, etc., like a physics and chemistry) The capacitance results from our ears hearing the relative proximities of notes nearing slightly less pure ratios.

For example a 5:4 third is rather odd when used in static tuning. It is considered more pure than the 12:ET tuned third, and it truly does sound like this. But if you play this interval in a triad on a synth - a pure 3:2 and pure 5:4 over the root note, it sounds like a standing wave. It sounds as if it is "too still". I'm all for understanding the math behind the more subtle melodic motions (and the capacitance that makes these motions so pleasing), which is obviously overwhelmingly difficult; more so by far than just validating Cope's interval roots theories since we need to distinguish the octave of the tones, and the infinite possibilities of tuning.

Sorry for the Wall of Words. I do understand that most people may not understand much of what I've said. But I feel blessed to meet another person who finds this field interesting as I do.

13 posts since 9 Jul, 2020

Post Sat Oct 24, 2020 3:26 am

Thank you for your message which is indeed extraordinary.
I think one has to be aware of the limits of the mathematical approach in order to be able to conceive theories that go much further than what can be calculated.
The examples you give show how a theory based on intervals and on the 12 chromatic notes can only come close to a very incomplete basis, from which one can go beyond the rule and go through the transfer of emotions, which is specific to the artist.

But I also think that classical harmony has locked us in a very rigid universe, which is not the absolute evil, because inside this closed universe, we master the rules better. Getting out of that is not reassuring, because we totally lose our references, so we would have to explore, risk not being able to get out of it, risk being crushed by a combinatorial explosion of possibilities, which frightens us very much.

In any case, thanks again.
And long live the theories,

User avatar
198 posts since 9 Aug, 2013 from The Hague, The Netherlands

Post Tue Nov 24, 2020 3:50 pm

Let the robot do it.
Don't give real musicians a chance in their effort.
"hah, a computer can make better music than you, son." ... otes/62253

63 posts since 8 Oct, 2019 from Lannion, France

Post Thu Nov 26, 2020 12:26 pm

Is this an arpeggiated post ?

50 posts since 24 Jan, 2018

Post Fri Dec 11, 2020 5:18 pm

There was those lawyer guys who tried to copywrite every chord progression ever lol. Automatic music is possible but i think it will always benefit from some finishing touches.

Return to “Music Theory”