How to Make Interesting Generative Music

[rofilmmedia]

After more than 40 years of working with modular synths I´ve got the idea of writing down my experiences and the knowledge that I´ve gained over the years and produce an e-book trilogy about how to make generative music. The first two volumes are finished, some hundreds of page, legions of videos and presets, and right now I´m working on volume 3 of the trilogy.

 

I think that generative music is an important part of working with modular synths, perhaps one of the most important ones. Therefore I´ve decided to make the main ideas and techniques of my work public – not everyone can afford to buy the books.

 

I´m going to rework volume 1 into a series of articles / posts and will even add most of the video documentation, and I hope you´ll find my posts interesting and perhaps they will initiate an in-depth discussion of the matter.

 

And that´s what I´m going to write about here (I´ll try to post one article per week):

(more about it all on my website https://dev.rofilm-media.net)

 

Enjoy your day!

Rolf

 

Chapter 0: About This Course And Some Words About

What Generative Music Is

 

Chapter 1: Real Randomness vs. Complex Cycles

(and the combination of both)

Chapter 1.1: LFOs

Chapter 1.2: Other Devices Generating Regular

Cycles

Chapter 1.2.1: Looping Envelopes

Chapter 1.2.2: Sequencers

Chapter 1.2.3: Shift Registers With Feedback

Chapter 1.2.4: Sequential Switches

Chapter 1.2.5: The Turing Machine – Part 1

Chapter 1.2.6: Samples and Recordings

Chapter 1.3: Randomness, Probability and

Stochastic

Chapter 1.3.1: Some Basic Definitions

Chapter 1.3.2: Sample & Hold

Chapter 1.3.3: A Short Glimpse at the Turing Machine

And at Shift Registers Again

Chapter 1.3.4: Perfect Pseudo Randomness:

Gray Code Modules

Chapter 1.3.5: Imperfect Pseudo Randomness:

Euclidean Sequencers

Chapter 1.3.6: Random Trigger (Percussion) Sequencers

with Different Amounts of Randomness

Chapter 1.3.7: Stochastic Sequencers

Chapter 1.3.8: Probability Gates

(Random Clocked Gates)

 

Chapter 1.3.9: Bernoulli Gates

 

Chapter 2: What to Modulate And to Trigger

Chapter 2.1: Pitch

Chapter 2.2: Timbre

Chapter 2.2.1: Filter

Chapter 2.2.2: Shapers

Chapter 2.2.3: Partials (additive)

Chapter 2.2.4: FM/PM

Chapter 2.3: Voices

Chapter 2.4: Rhythm

Chapter 2.5: Effects

Chapter 2.6: Envelopes

Chapter 2.7: Quantizers

Chapter 2.8: Grains

Chapter 2.9: Sample (Player)

Chapter 2.10: Slew Limiter

Chapter 2.11: Comparators

Chapter 2:12: Pitch Shifter

 

Chapter 3: Compositional Aspects of Generative Music

Chapter 3.1: General Thoughts, Strategies And

Basic Compositional Decisions

Chapter 3.2: Basic Compositional Techniques

Chapter 3.2.1: Contrasting

Chapter 3.2.2: Repeating, Modifying and

Inverting Relations

Chapter 3.2.3: Basic but Exclusively

Generative Techniques

 

 

Chapter 3.3: Specific Compositional Techniques

Chapter 3.3.1: Pitch Dependency

Chapter 3.3.2: Rhythm

Chapter 3.3.3: Tension and Layers

Chapter 3.4: Certain Patch Techniques

And Examples

Chapter 3.4.1: Switching Voices and Larger Parts

of the Patch

Chapter 3.4.2: Sculpture Randomness and

Setting Borders

Chapter 3.4.3: Jumping between certain BPM and

Inverting Pitch Lines

Chapter 3.4.4: Mixing Stable and Random Elements

 

Chapter 4: Some Building Blocks of Generative Patching

Chapter 4.1: The Instrumentation of Envelopes

Chapter 4.2: 5 Faces of Randomness

Chapter 4.3: Random Harmonies

 

[rofilmmedia]

Welcome to part 2 of this series of articles taken from the e-book (see https://dev.rofilm-media.net for some background information). Today we start patching, and I have integrated even video in this article to make things audible and visible.

 

Chapter 1:

Real Randomness vs. Complex Cycles

(and the combination of both)

 

Chapter 1.1:

LFOs

 

When we hear “permanently changing” most of us will surely think of sample and hold units at first.

And, yes, S&H units are important engines to drive our generative patches. But what about clock generators and LFOs (the latter being able to serve as clock generators as well)?

“Why, LFOs generate regularly repeating cycles?” you may say. And: “No permanently changes will be going on. All changes repeat exactly the same way, when the next LFO-cycle starts.”

You are true. Of course you are. But such LFO cycles can be quite long ones. The lowest frequency of the VCV rack LFO-1 for example is 0.0039 Hz, which means a cycle of 4 minutes and 16 seconds before things start repeating again. And there are LFOs with even lower frequencies and longer cycles out there. But even 4 minutes may give us – as listeners – at 

least the illusion of “permanently changing”.

Your next argument will be:

“But these changes are going on THAT slowly, that the result is boring at the least, and some of our listeners may even think, that there are no changes at all.”

And you are true again. But if my LFO is equipped with a CV-in jack to modulate its frequency, well, then things start to get interesting.

 

In other words: let´s talk about frequency modulating LFOs, about modulating the modulation strength (the “volume” of an LFO´s output), about feedback loops consisting only of LFOs, and about additive mixing of different LFO outputs.

https://www.dropbox.com/s/xhkhoqbqog7ncei/image 1_1 four ways.png?dl=0

You can imagine what complex networks we can build with these four building blocks.

 

And if we use different LFOs of such a network to modulate or trigger different sources of sound, we are able to construct a “super-cycle” (which consists of a set of “sub-cycles”) that lasts a very long time until it returns to its beginning.

 

And when we further take into account, that frequency is not the only parameter, which we can modulate, things get really exciting: modulating the LFO´s amplitude, the LFO´s phase and even the LFO´s wave shape (if our LFO is equipped with a CV in jack allowing us to modulate the shape). 

 

A simple example shall explain what I mean:

Let´s take two LFOs, LFO A and LFO B. LFO A runs at a frequency of 0.03 Hz, which is a cycle length of 33 seconds. LFO B runs at 0.04 Hz, which leads to a cycle length of 25 seconds.

LFO A modulate the frequency of a VCO, let´s call it “VCO X”, and LFO B modulates the frequency of a VCO called “VCO Y”. We can use two quantisers and two VCAs to make things more comfortable to hear and to listen to.

https://www.dropbox.com/s/oqoyvav1h416dir/image 1_2 patch.png?dl=0

 

Let´s now say, that both LFOs start their first cycle at the same time.

 

To use the aforementioned terms: we have one “super-cycle” consisting of two “sub-cycles”.

 

Please look at the following table now: it lasts all in all 825 seconds until both LFOs begin their cycles at the same time again. LFO A needs 25 cycles to get there, and LFO B needs 33 cycles.

The length of our “super-cycle” is 825 seconds, even if the “sub-cycles” are only 25 seconds and 33 seconds long.

 

The video “Video C1_1 SupercycleVideo” (just follow the link: https://youtu.be/Y0RFxQf3NZA ) demonstrates the patch.

Well then, let´s set up some typical LFO networks now.

https://www.dropbox.com/s/nphgdxkdttc8swb/image 1_3 cycles.png?dl=0

The easiest group of LFO networks – easy in terms of predictability – are additive ones, networks in which all LFOs work parallel on one and the same (ore more than one) modulation target. Let me start with 1 VCO as the target and 2 modulating LFOs. In those cases the absolute values of the LFOs add to each other: when the phase of the waves of both LFOs are positive the sum is a higher positive one, if both are negative the sum is a higher negative one, and if one wave is in its positive phase, the other in its negative phase, then we get a subtraction, and in case the waves differ by 180° we get a complete phase cancellation. The resulting summing wave may look like in the following picture (just an example). And patching a quantizer between the LFOs and the VCO we get the following melody (given that we have chosen the output strenght of the LFOs adequately (later more about adequate modulation strengths).

https://www.dropbox.com/s/1gqsd3jn7rnqehz/image 1_8 additive to notes.png?dl=0

 

In the region marked in yellow the patch will play

h2-#a2-a2-#g2-g-#f2-f2 (will be held for a while given, that the local minimum is still nearer to f2 than to e2) and then continue with #f2-g2-#f2-f2-e2-#d2-d2-#c2-c2#c2-d2-#d2-e2-f2 (will be held for a while again) and then back to e2-#d2-d2-#d2-e2-f2

 

With both LFOs running at different frequencies we get random sounding melodies with patches like that one.

 

Using a suitable mixer for the two LFOs we can adjust the wanted frequency ranges by adjusting the “volume” of the LFOs – what is adjusting the Cvs, which are sent by the LFOs. https://www.dropbox.com/s/rnpm77t75iorfx9/image 1_9 mixer_1.png?dl=0

And we can – of course – adjust the LFO output strength differently for both LFOs.

 

If the channels of our mixer are equipped with CVins, we can modulate the relative strength of each LFO by other modulation sources – e.g. using more LFOs. And if our mixer doesn´t have CV ins, we can patch VCAs between each LFO and the mixer.

https://www.dropbox.com/s/d34fbvs1i9c2c34/image 1_10 mixer_2.png?dl=0

But I´m anticipating what will be talked about later.

The following link leads you to a video, which shows me messing around with the patch (and explaining a bit more about it). In the video I´m using other LFO waves than only sine waves as well.

https://youtu.be/ZW3jcsFh0jg

 

Let me add a third LFO. But instead of just another sine wave this new LFO shall produce a square wave, a wave that simply jumps between two values. We will have to attenuate the output of this LFO a lot to avoid long periods of silence, when it´s at its low level.

https://www.dropbox.com/s/xnmhv3fimwdrz3n/image 1_11 mixer_3.png?dl=0

 

With a patch like that we get “unexpected” jumps in the melody, which had been simply going up and down so far – continuously up and down – and the only “randomness” was in the different lengths of the rises and falls.

And with this third LFO, which – or course – runs at a third different frequency, the overall length of the aforementioned “super-cycle” increases dramatically, which increases the impression of randomness even more. With frequencies of 0.025 Hz, 0.035 Hz and 0.25 Hz our “super-cycle” gets a length of 1,160 seconds, what is nearly 20 minutes. Surely long enough to cause the impression of real randomness.

The video, which is hiding behind the following link demonstrates (and explains) the patch in detail.

https://youtu.be/i82oluhKj7M

 

In the next article I´ll leave the field of purely additive combined LFOs, build up some rather complex networks and introduce a general LFO block system to make it easier to construct and to document LFO networks of infinite complexity.

 

… to be continued