Specified in the description of this January 21, 2021 video.

Getting somewhere with my self-playing patch goals - no samples, all synthesis! I'm enjoying this patch for its breadth of dynamics and sonic variation. There's a lot going on here but I'll share the main points and major principles:

Mutable Instruments Marbles provides the main clock. An ALM Busy Circuits Boss Bow Tie randomly switches between the negative and positive LFO output of a slowly cycling Make Noise Function, and that signal modulates the clock rate - sometimes launching it into audio rate or slowing it to a crawl. At times the clock rate remains steady when Function's Hang input is engaged by a gate from a 4ms Quad Clock Distributor + Expander, which itself is being clocked by Marbles.

Sacrament Cursible switches between six sound sources: Synth Tech E330, After Later Juggler, two channels from Endorphin.es Furthrrrr Generator (one going into Intellijel Rainmaker), Mutable Instruments Ripples self-oscillating into Monsoon (left and right channels). Each voice has varying degrees of random modulation dialed in.

Cursible's output is fed into a VCA opened by a Make Noise Contour. This is the only envelope in the entire patch! Its attack and decay are both being randomly modulated but with a bias toward shorter gestures. An interesting chain triggers it: Marbles' t3 output, fed through Branches for random skips, clocking a 4ms Pingable Envelope Generator with its clock division being randomly modulated, and taking the Half-R gate output from there.

The signal is then multed to both sides of a 4ms Dual Looping Delay, both at half wet, with their delay times being modulated independently. This results in some of the glitchy sounds you hear.

Those two signals are then sent through different but mirrored chains - a modulated high pass filter followed by a compressor. In one case it's Random*Source Serge VCFQ into Audio Damage Kompressor, and in the other it's Ritual Electronics Altar into 2hp Comp. Both filters' frequency and resonance parameters are being modulated with smooth random. The signals are then mixed with Make Noise X-Pan, with a little modulation over pan amounts.

X-Pan's outputs go to a Make Noise Mimeophon, mix at about 40% wet. Skew is on so the two channels' delay times differ from each other, and most parameters are being modulated so that delay times might be tiny or huge and have varying levels of feedback. This is also the reason why multiple sounds are often layered in the patch even though Cursible is only spitting out one at a time. Halo (Mimeophon's reverb) is swelling in and out with semi-random timing.

Modulation sources not mentioned above: Instruo DivKid OCHD, ADDAC VC Transitions, Vermona Fourmulator, Qu-Bit Bloom, NLC Triple Sloths, Moffenzeef Deviant, SSF DivKid RND STEP, Doepfer A-148, Make Noise Maths, Make Noise Telharmonic, 2hp LFO.