The ever-changing generative patch challenge?

Looking for good approaches of creating ever-changing generative patches, i.e. those that sound and feel entirely different after letting them run for 5, 10, 20 or 30 minutes? They slowly but steadily change and morph into something new when leaving them unattended.

Still would want sounds to be pleasing to listen to, but am trying to escape long-term repetition in any way.

In other words: A patch you could listen to for hours without ever feeling bored.

If you may, please post your example patch(es) here :smiley:.


there are maybe some interesting examples in “redmeansrecordings” series of “VCV for sleep…” patches,

you can download some here:

and listen to them on his youtube channel:


I find that slow modulation goes a long way in creating ever-changing patches. Recently, we were lucky enough to get Nonlinearcircuits’ Sloth Chaos modules ported to the library: Torpor, Apathy, Inertia and Triple Sloths. From the manual:

Torpor takes about 15-30 seconds to travel around 2 strange attractors. Apathy takes 60-90 seconds and Inertia takes 30-40 minutes. The behavior of these circuits is unpredictable over time, but more orderly than random. They are often used for low frequency control voltages in cases where an LFO would be too predictable, but a random signal would be too erratic.

Those are well worth checking out. Other than slow modulation, for the purpose of your question I’ve also used feedback patching (I can recommend La Synthèse Humaine’s YouTube series on the topic of cybernetics, as well as Simon Hutchinson’s videos on artificial neurons and nonlinear mixing) and binary counters (to output gates which were later converted to triggers and voltage).


Thank you much for the replies!

The Red Means Recording sleep videos and patches are really useful and carry a number good ideas for introducing variation.

I also think that slow modulation can reach very far. The Sloth modules are great and remind me a bit of Vult’s Caudal, which I use a lot. Caudal adds speed control, and an energy setting, and has a couple of different models, but it doesn’t make it down to the slowness of Inertia. And Triple Sloth adds another dimension, so they all are good additions to the toolbox.

Using feedback and cybernetics is an interesting but also somewhat challenging approach. It seems to take some skill to control feedback stability. I suppose I’ll just have to keep in mind not to blow my speakers when the frequency escalates into the audio range.

I am also exploring how to automatically record segments of audio while the patch is running and then re-play them in altered form, e.g. after being pitched down. Obtained some first results with NYSTHI’s Simpliciter.

Moreover, dialing voices in and out at either defined or at random times may yield intriguing combinations.

1 Like

the beauty itself of a modular universe: so many ways to do this.

my approach is to modulate everything in a range that sounds pleasant to my ears, MAYBE going slightly outside this ranges sometimes. random generators, scale modules and probability triggers are my best friends.

choose a mood and put a lot of reverb.

this was my first experiment (the first one I liked to be honest, and that’s a single note piece, not even modulated)


something more “not-so-nowadays” - because it can be done lighter on the cpu

1 Like

These are very important tools for sure.

Your track is very peaceful, I like it!

Still, my current thought is to go beyond the regular generative ambient by very slowly adding more and more variation in the course of the track (on the scale of minutes or more), trying to form segments that on one hand exist in their own right, but on the other hand still correspond with each other somehow, even if it happens in a (semi-) random way.

Perhaps an easier way to steer toward this would be to record different patches and splice the audio tracks together. But my idea here is to combine everything in one patch to maintain an element of surprise and morph them more fluently.

1 Like

if i’m allowed to toot my own horn, we already had wiqid anomalies which includes languor, which is heavily inspired by triple sloths.


I know what you mean, but it’s also difficult to know whether you left anything out. But just taking you at your word: The sound and feel of a piece is determined by tempo, rhythm, timbre and melody. And also of course factors such as how simple or busy it is (#of voices) and choices of note scales. So those are all the things you would want to modulate to create constant change. The hardest part of what you say is actually avoiding long-term repetition, as most modulation tend to be somewhat cyclical.

Tempo and rhythm are not difficult to modulate, although a pleasing rhythm at all times can take some work. Maybe look at a module like Grids/Topograph.

Modulation of timbre can be accomplished in countless ways, typically through mixing and modulation of oscillators, filters and effects.

Ever changing and pleasing melody is the hardest part I think. You want a quantizer and/or sequencer you can modulate, that doesn’t get out of the pleasing bounds. Something like the teletype module is a powerhouse in that respect, but also has a learning curve. I can highly recommend the module NYSTHI:Bitshifter through an advanced quantizer such as those from Frozen wasteland or Harmonaig, but we have an abundance of sophisticated quantizers. If you don’t mind paying for things the ML:Quantum Multiverse would probably be very good for the job. To make melody changes very gradually I can highly recommend the brilliant module ML:Evolution. There are also sofisticated melody generators, or chord generators/modifiers such as Meander.

For modulation itself you want the usual gallery of participants: Sample & Hold, chaotic LFOs/smooth random, very slow LFOs, etc. We have lots and lots of those.

Here’s a first minimal template for something like this, quick-and-dirty, only one voice, not very musical, etc. You’ll have to add many more modulators of different speeds, and voices and make it suit your taste. But it illustrates the principles mentioned.

Generative.vcv (18.1 KB)


the new VCV module RSCL (rescale) is super helpful for a generative patch. finding the correct range of a random source is a key. before I used BARK CLAMP and NYSTHI SCALE (they are still in many of my patches and will use them in the future of course), the new VCV module let us reflect the value as well, both at max and mix value. next stuff is about mixers: just mix values with tiny random voltages, everywhere :sunglasses:

1 Like

Thank you for all these really useful tips! I will have a lot to try out :smile:. Your patch example is very nice, too!

Recap and thoughts:

  • Use modulation of: tempo, rhythm, timbre, melody, # of voices, key/scale, and probably others

  • Try Valley’s Topograph for varying the rhythms

  • Timbre can be modulated with oscillators (LFOs), filters and FX, among others. Lots to choose from

  • Modulate quantizers and sequencers. Teletype is steep indeed, but will take a look

  • The NYSTHI Bitshifter (BOH!NGLER) is 256-bit shift register. Found a decent manual for it (sometimes a bit difficult with NYSTHI modules)

  • ML’s Quantum Multiverse is similar to 8 instances of Quantum (ML). Will start out with this one

  • Frozen Wasteland’s Probably Not(e) is a probabilistic quantizer that looks quite interesting, as does the diatonic harmonic quantizer harmonàig by Instruō. Will take a look

  • ML’s Evolution, a sequencer based on shift register, is another good one. Plenty of options

  • PurrSoftware’s Meander will take some time to understand better, but fortunately there is an extensive manual and a couple of tutorials on YT

  • For modulation: Sample & Hold, chaotic LFOs/smooth random, very slow LFOs, etc

One thing I believe would be helpful is to master ways of switching between two or more signal paths (all voltage, so i.e. CV, clock, audio etc.) either at random or defined times (e.g. clock counters).

Of course, should stretch out the timeline for everything and let changes flow in small steps to keep the morphing smooth.

1 Like

RSCL is wonderful (even though my mind always wants to spell it out as ‘rascal’ :laughing:).

I also like using Fence by OrangeLine to curb the upper and lower limits of a signal, including pitch but also for voltage/CV.

For the long-term motion I am seeking, I find it especially useful that the limits themselves can be modified by CV, and that would be a good addition in RSCL.

Never used Bark Clamp, so will take a look.

Mixing moving CVs with random voltages is an excellent idea to introduce small scale variation! Need to keep range adjustment in mind, if necessary.

1 Like

I’ve tried using more than one Seriously Slow LFO’s set to different numbers of minutes to control whatever you want to slowly change:

1 Like

Being able to modulate a signal with SSLFO over minutes and perhaps hours is really helpful.

Is anyone using the longer timebase settings (weeks and months), too? It seems that even over extended run times of a patch (let’s say 6 hours would be reasonable), they generate more or less flat lines. So, do they have a practical use or were they included just for fun?

BTW, Seriously Slow Envelope Generator also is a good companion for slow motion.

As for so many things VCV Rack and Modular, @Omri_Cohen also has made and shared many informative, inspirational and step-by-step videos on Generative patching tools and techniques.

The link below is a search for keyword ‘Generative’ on Omri’s YouTube channel (with many results) Omri Cohen - YouTube

1 Like

that would be a joy for me: it would mean that somebody wanted me to manage his (multichannel) exibition :sunglasses:

1 Like

That is an excellent idea! IMO, many of @Omri_Cohen’s videos are absolutely top-notch and I watched many. Perhaps he’s interested in a tutorial on key concepts of the ever-changing patch.

Haha, that would be perfectly secure job for a lifetime :laughing:.

Collecting additional concepts while researching and learning.

Trigger sequences that speed up and slow down over time can be created easily. Resulting sequences follow a somewhat more complex structure, but are not random.

a.) Rampage by Befaco: Modulating fall and/or rise of envelopes A and B control the shape of the trigger patterns

b.) FM with two LFOs (e.g. LLFO by Bogaudio): LFO1 modulates frequency of LFO2. Can use Scale, Offset and Frequency of LFO1 to vary the trigger pattern.

LLFO (Bogaudio)_Speeding up and slowing down trigger patterns

1 Like

Some random tools and tips for use in Generative Patches…

Sample & Hold is very common and uselfull in Generative Patches to, well…sample and hold…voltages when triggered. Voltages (CV) could represent anything in modular. E.g. clock/tempo, pitch, amplitude etc.

If more gradual transitions between values are needed, there’s the Slew limiter. Can also transform (sharply edged) gates or triggers to more gradual envelopes. This averageing/dampening effect is not allowing fast/large transitions between sample values. So, effectively not allowing higher frequencies, so effectively a crude low pass filter (yes, also for audiorate input).

The Bernoulli gate adds probability to gating/triggering.

A Flip/Flop can be usefull to gate stuff (e.g. switch stuff on of off) until the next gate/trigger. Effectively a Filp/Flop is also a Clock Divider (/2).

An Edge Detector can gate stuff depending on the current slope of a signal at the input (rising/falling).

Underrated: the fixed Voltage and Offset. Also very usefull in Logic.

A Comparator compares Voltages and generates Gates (or Triggers) to do stuff as soon as some relative condition is met. A Comparator generally takes 2 inputs and applies logic. E.g >, >=, =, <=, <. The binary output can be used as a Gate. Such a gate holds as long as the logic is true. Some can also output a Trigger (which will not hold).

Less common is the Accumulator. You can use an Accumulator to manage voltages over time. Basically, if an Accumulator receives a Trigger/Gate, then it adds a voltage to the voltage it holds at that point in time (drops in a bucket). Either a set/fixed voltage or whatever value it reads from an input. Negative voltages will subtract. The Reset can be used to revert back to 0V (empty the bucket).

You can combine Accumulator with a Comparator to open a Gate as soon is some relative condition is met. Or use a Comparator to trigger an Accumulator as soon as some relative condition is met. Combining with Accumulator Reset can create a cycling level/counter where the Comparator resets the Accumulator as soon as some level is reached.

The DocB ACC Accumulator is pretty versatile.