It all started with watching a video of a hardware Eurorack patch that uses a number of Mutable Instrument modules (the video in question is Tundra by ‪@SyllixMusic‬). I tried to recreate the patch in VCV Rack to attempt to figure out how each module contributed to the overall result. I got nowhere near the same result, both because the Beads module doesn't have a VCV Rack counterpart (yet?) and because I didn't get the correct settings for the other modules in the short time that I invested in it. It did however lead to some interesting sounds coming from the Macro Oscillator module (the VCV counterpart of the Plaits module) when set to the "Granular format oscillator" model, especially when playing with the Timbre CV, the Morph CV and the Lowpass gate decay.
Those knobs can't be turned directly using CV power (and I wanted the CV inputs for Timbre and Morph unpatched so that they remain linked to the internal envelope generator), so the stoermelder µMAP module is used to automate the knobs (with the Instruo Ochd module and a Sample and Hold with a 4 second glide configured).
For note input into the Macro Oscillator, I turned to one of my favourite ways of generating an interesting main voice: merge multiple clocks (a quarter, eight and triplet clock in this case) into a polyphonic signal and feed that into the Impromptu ProbKey module. Just reduce the density value (so that not every clock triggers a note) and increase Lock value (so that a slowly evolving sequence is constructed by the module). The notes are set to a C minor scale, with a higher probability on the notes of the pentatonic minor.

As usual, this main voice gets supported by a bass line. The AddrSeq module (with he ASX extension module) has four sets of four bass notes programmed, again in the C minor scale. Each time one sequence ends, the next sequence is selected at random by a Sample and Hold module into a 8:1 switch. The resulting bass notes are sent into two Dark Energy modules (set into two octaves and modulated by a Walk2 module) and a pair of Basal modules (which are modulated by the Ochd module).

That's what the patch consisted of for a couple of days. A number of attempts were made to add additional voices to this, but none sounded right in the mix. The only changes that were kept in this time was some tinkering with settings of the earlier described voices, and some additional effects on the main voice. One addition there was to add a pingpong delay (using the Chronoblob2 module) into a separate channel on the mixer. This adds a slowly decaying tail on the main voice. Another late addition to the main voice was to add a reverb module before sending it to the mixer, making it a bit wider (which was a real "omg" moment when I plugged it in). By modulating the width and mix amount of the reverb, some additional movement is also introduced here.

After a couple days of tinkering with the patch every now and again, I finally figured out why my attempts at adding additional voices had failed so far: the patch didn't need that much added to it anymore. The main voice is already taking up a lot of space and adding something with a bit of complexity into the mix kept causing a battle for attention.
So I set out to add a more simple voice. Some long notes slowly fading in and out in the high register. Through the dIKT@T and Collapse module, a random note is picked from the 1st, 3rd, 4th, 5th and 7th notes above the currently active bass note, and that is sent into a VCO. The (filtered) triangle, sawtooth and sinus outputs of the VCO are mixed together (modulated by a 4FO module and overall volume controlled by an ADSR envelope) and the result is sent into a Tap Dancer delay. The bonus provided by the Tap Dancer is that once the output of the VCO is muted by the ADSR envelope, only the delay output remains audible. And since a modulation effect is applied to that delay, this causes the note to slightly fluctuate in tone as it fades out. It was a delicate balance to get right, since to much fluctuation in tone just made it sound over the top. Now you barely notice it is there, but disabling it on the module immediately reduces the depth of the voice.

One final addition to the patch is some noise that is sent through a filter and a VCA, both of which are modulated by a Sample and Hold module with a six second glide applied to it. Two reasons for adding this noise: (a) it provides a nice atmospheric background with some movement and (b) there was some space left in the patch once I started moving all the modules into a rectangular shape.

__ Setup __
VCV Rack 2.3.0 on Windows
All audio generated internally within VCV Rack (no external connections)
Video is captured by OBS
Audio output is sent through a VBAudio Virtual Cable to Ableton Live
Audio recording synced to OBS, and fade added at the end with Reaper
OBS video and Reaper audio are merged and cut to length using Shutter Encoder.

__ Copyright __
All rights reserved