Software

d14spora.studio

34121

D14SPORA is a digital studio designed and developed by MNNN as a didactic tool for the Department of Sound at the Kunsthochschule Kassel.

Part performance system, part autonomous sonic environment, it unfolds through an ensemble of independent voices. Each voice can act as a living agent: listening, generating, responding, and finding its own path through a shared sonic field. Nothing is prerecorded. Nothing is fixed. Each voice develops its own tendencies and sense of time, guided by a common root and scale.

From these conditions, something unstable begins to take shape. Voices intersect, diverge, resonate, and fall away. No composition is predetermined; structure emerges through their ongoing encounters and interactions.

D14SPORA is a system in continual becoming: A digital space where temporal events relate and evolve, and where complex forms are not imposed, but discovered.

STRUCTURE

The ensemble is divided into three main collections.

VOICES

168 voices arranged as twelve banks of fourteen. D14SPORA's primary sound generators.

STRUCK AND RESONANT BODIES
A bank of tuned modes - a ratio, a level, and a ring time for each - set going by a burst a few milliseconds long. The strike decides only how the sound begins; the body decides what it is, which is why one excitation makes a bell, a bowl or a plate depending on what it is thrown at.

PHYSICAL MODELS
The instrument as a delay line rather than as a spectrum. A plucked string is noise sent round a loop the length of the string and filtered a little on every lap; the mesh is four such loops meeting at a junction, scattering one strike between them until it settles. Nothing here is a recording of anything.

GRANULAR PLAYBACK
Grains read from buffers the instrument records into itself while it loads - five seconds apiece, its own sound rather than a library's. Density, position, rate, and window are separate controls, so the same five seconds is a texture, a stutter, or a held chord depending on where they sit.

SPECTRAL VOICES
The sound is taken apart into two thousand bins, altered as a spectrum, and put back. Bins are scrambled, frozen, thinned to their peaks, or moved; what comes out keeps the material's colour and loses its behaviour, which is what makes a freeze sound like a room rather than like a loop.

CHAOTIC OSCILLATORS
Deterministic systems that never repeat - a Lorenz attractor, a sine folded back through its own output. Not noise: the shape is stable, and the detail is never twice the same, which is drift at low rates and a source in its own right at high ones.

ELASTIC, VOCAL, HARSH, ELECTRIC, IRREGULAR, AND ORGANIC TIMBRES
Not a sixth kind of voice but the ground the five above cover between them. Each is built to reach several of these rather than to hold one, which is why moving a single fader crosses from one to another instead of giving more of what was already there.

HARMONICS

168 voices arranged as twelve banks of fourteen.

Voices tuned to integer multiples of the global root rather than to the active scale. Because every pitch is derived from the harmonic series, the prism remains internally consonant even as the musical scale changes.

The voices include just-intonation intervals, microtonal relationships beyond equal temperament, and spectral treatments derived directly from the harmonic series.

PULSES

70 voices driven by an independent sixteen-step sequencer, the last twelve of them a bass section.

Unlike the free-running families, pulse voices are synchronised to their own clock. Four voices replay the external input rather than synthesising sound directly, allowing live material to participate in the rhythmic layer.

5P3C5

VOICE GROUPS

D14SPORA is organised into three independent voice groups. Each group employs a distinct method of sound synthesis while sharing the same global temporal field.

HARMONICS

Rather than following the active scale, harmonic voices are tuned to integer multiples of the global root frequency. This preserves consonant relationships throughout harmonic changes, regardless of the selected scale.

JUST INTONATION
Intervals as whole-number ratios rather than as equal steps: the triad is 4:5:6, and tuned that way its thirds beat with nothing at all. The stillness is only audible while two voices hold the interval, so it wants sustain and few notes.

SEPTIMAL AND UNDECIMAL INTERVALS
Dyads on ratios no equal temperament can reach. Seven over four is the flat seventh that makes something sound archaic rather than modal; eleven over eight falls halfway between a fourth and a fifth and belongs to neither, and that ambiguity is the whole of it.

UNDERTONE SERIES
The harmonic series the other way up: the root divided rather than multiplied, so the partials fall below it instead of above. The same arithmetic downwards, and what it builds is a chord under the note rather than a colour on top of it.

PYTHAGOREAN STACKS
Pure fifths piled one on the next until they leave the octave. Nothing is tempered to make them fit, so the thirds arrive wide and restless - what western music used before it wanted chords - and under a long stretch they beat slowly enough to be heard as movement rather than as error.

BOHLEN-PIERCE STRUCTURES
Built on a twelfth rather than an octave and divided into thirteen, so nothing in it is an octave apart. The prism's harmonic layout disagrees with it entirely, which is worth hearing once.

BEATING PAIRS
Two partials sounding together a fraction of a hertz to twelve hertz apart. Neither is heard as a pitch of its own; what is heard is the difference between them, the amplitude swelling and falling as many times a second as there are hertz in the gap.

DIFFERENCE TONES
Two high partials spaced so the difference between them lands back on the root - seven times it and eight times it, whose difference is it. The root is played quietly underneath as well, so the tone the ear constructs and the tone that is actually there arrive together.

PULSES
Pulse voices operate on their own clock, independent of the global temporal field. Instead of freely generating phrases, they derive rhythmic structures from the sequencer while maintaining independent pattern lengths, rotations, and timing relationships.

Several rows share the same synthesis engine at different fundamental frequencies - for example, two kick voices or four sub-bass voices - allowing related sounds to interact as a coherent rhythmic system rather than as isolated instruments.

DYNAMIC STRUCTURE

D14SPORA can operate without user intervention. A dedicated structure engine continuously reshapes the musical state by adjusting its own parameters, enabling and disabling voices, changing scale and root, transposing the harmonic field, rerecording its source material, and recalling previously stored states.

Rather than following a predetermined sequence, the studio engine can gradually reorganise the ensemble over time, producing long-form evolution while preserving the independence of each voice.

PRESETS

A preset captures the complete state of the system, including all continuous control parameters. Stored presets may be recalled at any time, allowing the structure engine to revisit earlier configurations as part of its autonomous behaviour.

Transitions between states are performed by interpolation. During a morph, continuous parameters are updated at 25 frames per second, producing smooth changes throughout the studio engine.

Discrete parameters are excluded from interpolation. Voice activation states are switched directly rather than blended, and the sequencer grid is stored and recalled independently of the continuous parameter set.

VOICE ARCHITECTURE

Every voice consists of three independent components.

SYNTHDEF
Produces the sound itself. One synthesiser instance is created for every note that the voice generates.

PDEF
Generates musical events. The pattern determines when notes occur and supplies parameters such as pitch, duration, articulation, register, and timbre.

CONTROL BUSES
Continuously changing values read by active synths. Pattern values are evaluated only once, when a note begins.

Continuous controls are therefore read directly inside the running SynthDef from dedicated control buses. Parameters including level, energy, chaos, and effect sends remain responsive throughout the lifetime of every note.

Changing the level of a sustained voice affects the note already playing rather than waiting for the next phrase.

INTERNAL BUILDS

Every voice contains three internal constructions. Oscillator topology, excitation model, resonator configuration, or modal ratios, depending on the voice.

One build is selected for each phrase and remains active until that phrase has completed. The result is that a single voice can produce substantially different timbres without changing its musical behaviour.

VOICE CONTROLS

Opening a voice exposes multiple parameters dedicated to that voice.

RATE
How often this voice speaks, from a tenth of the field's pace to eight times it.

PHRASE LENGTH
How long each phrase runs, on the same scale as the rate: a tenth of the field's measure to eight times it.

SEND LEVEL
How much of this voice reaches the sections its family sends to, up to three times unity.

BUILD SELECTION
Which of the three builds this voice uses. Below zero, it follows whatever the family is set to.

SPECTRAL CONTROLS

RESONATOR CONFIGURATION
How many modes the bank holds and how they are spaced. The same excitation through a different configuration is a different object, which is why a bank is a family of instruments rather than one with a tone control.

PARTIAL DISTRIBUTION
Where the partials sit above the fundamental. Whole multiples give a pitch; anything else gives a body that rings without settling on a note.

RESONANCE
Resonance at the cutoff. Up, the filter rings and the module gains a pitch of its own beside the carrier's.

SPECTRAL BALANCE
Which end of the bank is loudest. It decides whether a struck thing reads as wood, metal or glass without changing a single frequency in it.

FILTER CONTROLS

FILTER TYPE
Which of the fifty filters the voice is played through. Each opens its own panel, so several can be set up and left, and the choice is a character rather than a setting.

CUTOFF
A low-pass after the folding, in hertz.

RESONANCE
Resonance at the cutoff. Up, the filter rings and the module gains a pitch of its own beside the carrier's.

DRIVE
Saturates this voice alone, before anything downstream. The gain is compensated, so it thickens rather than simply getting louder - it decides how hard the signal arrives rather than how much of the result is mixed back in.

Pulse voices provide an additional eleven parameters governing sequencer behaviour, including rhythmic generation, rotation, probability, and timing.

RESOURCE ALLOCATION

Resonators and filters are created only when required. A voice whose filter remains at its neutral position allocates no filter resources. Likewise, resonators exist only while their controls depart from zero. This dynamic allocation avoids maintaining thousands of permanently running processors.

A permanently allocated resonator for every voice would require approximately 2,800 simultaneous filters.

SIGNAL PATH

D14SPORA transforms digital audio through a succession of interconnected processing stages.

SOURCES
The six families that make sound: the external input, playback, voices, harmonics, pulses, and the FM module. Each has its own level and its own sends.

FILTER
Fifty filters, each a body rather than a setting. The section's cutoff and resonance are the gesture; each filter offsets them and adds its own character.

SKEW
The distortions: fuzz, germanium, crush, decimate, wavefold, tube, diode. Separate bodies, because the difference between them is not a number.

GATE
A rhythmic cut on its own clock — depth, rate, and the edge of each step, from a click at the bottom to a tremolo at the top.

UNREST
A delay line with a feedback loop and four filters — what a filter does to pitch, done to time. A smear at one end, a stutter at the other.

CAVITY
A bank of tuned resonators that answers what is sent to it with its own pitches rather than the material's. Rings on after the source has stopped.

LAB
The lab's one way in. Every family can send a copy of itself here, and this is what those copies arrive at before any of the nine sections below act on them — so it is the level everything else in this drawer is working from.

STRETCH (ONE PER FAMILY)
The same machine at every family, reading behind the write head, so a family can be heard stretched, held, or reversed beside its own direct path. The harmonics family has a prism here instead.

MASTER STRETCH
The master's own stretch. The same machine the families have, listening to the finished mix rather than to one send.

BUS
The compressor across the whole mix, and the headroom before it. It replaces the mix rather than adding to it, so it never stops running.

OUTPUT
The last stage before the card: colour, width, headroom, compression, and a limiter, per channel.

The harmonics family's stretch is a prism instead, with its own reverb after it.

Nine stages take a send from each family. The routing grid lists them in the order they appear in a family's panel.

LAB
How much of the external input goes to the lab. Its own, rather than the one every voice reads - which is why the input arrived there whether or not it was wanted.

FILTER
How much of the external input goes to the filter section. It is a separate path, so direct at zero and this one up gives the filtered signal alone.

SKEW
How much of the external input goes to the skew section.

GATE
How much of the external input goes to the gate. Because it is a send and not an insert, the input can be chopped on the gate's own clock while everything else in the ensemble runs on undisturbed.

UNREST
How much of the file goes to the unrest module - the delay that stretches and contracts what it is given. Nothing is taken from the direct path: this is a copy, and what comes back arrives beside it.

STRETCH
How much of the file goes to its own stretch.

PRISM
How much of this module goes to its own prism. A send of its own, so the prism can be used without the stretch and the other way round.

YIELD
Pulls the tail down while the input is loud and lets it back when it stops, so a dense passage stays legible, and the reverb is heard in the gaps. It's the last of them and the send that leaves the family for the master; a family whose yield is shut is audible only through whatever else it was sending to.

Each stage consists of one or more synthesis nodes that read from the audio bus written by the preceding stage. In addition to the main signal path, voices may send audio to a shared room bus and to dedicated cavity buses, allowing individual resonators to operate independently of the main processing chain.

The processing order is intentional. The recorder precedes the stretch processor so that the recording buffer always contains the original signal rather than the stretched output. Reversing this order would cause the recorder to capture its own processed output, creating an unstable feedback loop.

The master stage performs four operations in a fixed order:

Replace invalid floating-point values (NaN and +/-inf).
Limit sample values to the range +/-8.
Apply the master output gain.
Apply compression and final limiting.

This order is essential. Invalid floating-point values cannot be removed simply by multiplying them by zero; they must first be replaced with valid samples. Likewise, limiting is applied after the master gain so that the final output level reflects the actual signal presented to the limiter.

STRETCH

There is one stretch chain per family: voices, harmonics, pulses, external, playback, and FM. Each has its own send bus, its own eight-second buffer, its own recorder and its own player, so the material never mixes before it is stretched, and `stretch` on the pulses is the pulses.

Each family carries its own full set of handles, twenty-five of them, in its own panel.

LEVEL
How much of this family's stretch is heard, besides its direct path and its other sends. Without it, the separation is only in the buffer: the family could be stretched on its own but not heard at a level of its own.

TIMES
How long each mode rings. Short is a knock and long is a bell, and the spread between the shortest and the longest is what makes a body sound struck rather than played.

DIRECTION
Forward, frozen, or backward. In the middle, the read head stops, and the grains keep overlapping at that one spot, which is a moment held open.

BACK
How far behind the write head the stretch reads, up to about seven seconds. With hold, it becomes a scrub through what the family has just done.

PITCH
The pitch of the file, without touching its speed. Two octaves either way. Separating the two needs grains - the dry path is a tape, where speed and pitch are one thing - so moving this brings in as much grain as it takes. `grain` decides the character above that; at unison this does nothing and the file is read straight.

GRAIN
At 0, the file is read straight from the buffer - no grains, nothing added, and speed and pitch are one thing, as they are on tape. Above 0, it is granulated, and the two come apart: how long each grain is. Short is a texture and loses the file's own rhythm; long keeps it and smears the transients less.

BLUR
Scatters the grains in time so they stop landing on a grid. Without it, a long window repeats audibly at the window's own rate.

HEADS
How many playback heads read the buffer at once, one to four. One is a tape; four is a cloud, and everything between is how thick the stretch becomes.

SPAN
How far apart those heads sit in the eight seconds. Small keeps them together as a thickening; large has them reading different moments at the same time.

SPRAY
Moves each head a little on its own, in position and in pitch. This is what stops several heads sounding like one loud one.

SPREAD
How much of the file the scale is spread across. Narrow and it is one region read at different pitches; wide, and each degree of the scale is its own material, so the file is an instrument with twelve different sounds in it.

TILT
What the stretched signal keeps: lows below the middle, highs above. A stretch is mostly sustain, and this decides whether that sustain sits under the mix or over it.

FEED
Feeds the stretch back into itself, so what it produces is stretched again. Near the top, it stops being an effect on a source and becomes a drift of its own.

FEED TONE
What survives a lap of the feedback. Below the middle, the loop keeps the low end and darkens with every pass; above it, the loop keeps the top and thins. It is the difference between a feedback that drifts somewhere and one that only grows.

SNAP
How sharp the attack of every pulse voice is. Low is soft and rounded; high is a click on the front of the sound.

WARP
Bends the read pointer itself, so pitch and time bend together the way tape does and a pitch shifter cannot. Low is wow; high enough and the rate crosses zero, so the head reverses inside a grain and overshoots coming back.

WARP RATE
How fast the bend runs, from once every fifty seconds to thirty times a second. Above about ten, it stops being a warp and becomes a sideband.

WARP SHAPE
Sine at one end, triangle in the middle, stepped random at the other: a smooth waver, or a machine losing and regaining its grip.

JUMP
Throws the read head somewhere else in the thirty seconds, at the rate below. With snap up it is a hard edit between two moments of what the ensemble has just done; with snap down the same trigger becomes a slow sweep to the new place.

JUMP RATE
How often the head jumps, and how often each head rolls for its chance. Below one, it is a cut every few seconds; at the top it stops being a cut and becomes the grain rate itself.

JUMP CLOCK
Below the middle, the jumps run on their own clock, at the rate above, unrelated to anything else playing. Above it, they land on the sequencer's steps instead - every step, every other, every fourth, eighth, or once a bar as this goes up. The same mechanism stops interrupting and starts editing in time.

JUMP GRID
At zero, the head lands anywhere in the thirty seconds and never twice in the same place. Turned up, it lands on a grid of sixteenths, so the same moments come round again - which is the difference between scattering and composing.

CHANCE
How often each head speaks at all. At one, they all sound, which is how this has always behaved. Lower it and heads drop in and out on their own clocks - the cloud becomes pointillist, sparse and loud where it lands, which is the dynamic range this module has never had.

HOLD
Freezes the spectrum of the whole mix in bursts.

SEND ON (INTO THE MASTER)

SNAP
Scales every smoothing time in the chain at once. Every handle was lagged between a third and half a second, which is why nothing in the stretch could ever be sudden and why a modulator pointed at one arrived flattened. At the top, they step instead of sliding.

ELASTIC
Bends the read pointer itself, so pitch and time bend together the way tape does and a pitch shifter cannot. At depth, the rate crosses zero, and the head reverses inside a grain.

JUMP
Throws the read head elsewhere in the buffer on a trigger, either on its own clock or on the sequencer's steps, and either anywhere or on a grid of sixteenths so the same moments return.

A seventh chain reads what the six send it, with a thirty-second buffer: a stretch of already-stretched material, and a freeze there holds a moment of the whole instrument rather than of one family. Each chain has a compressor of its own with a four-band equaliser, because grains from different families have different peaks and evening them together would be the mixing this exists to undo.

UNREST

The fifth stage in the routing map, and the one the specifications above never mention. Every family reaches it by a send of its own - voices, harmonics, pulses, playback and FM - and the send is pre-fader, so a family sitting at zero still feeds it.

It is a delay line whose length is moving. That is a pitch shifter with a memory: the material sags as the line grows and snaps as it shrinks, and the shift and the echo are the same mechanism rather than two effects in a row.

LENGTH
Runs from eight milliseconds to nine tenths of a second - a resonance at one end, a repeat you can count at the other. `rubber` is how far that length is pulled about, and it is the whole effect: at zero the module is an ordinary delay, and everything above bends pitch by changing the distance the sound has to travel. `snap` is how quickly the length arrives where it is going, which is the difference between rubber and tape.

STEP
Transposes each pass through the feedback again. The middle is no shift; a fifth either way is a ladder that climbs or falls for as long as the feedback holds, and a few cents is a detune that never settles. Nothing acoustic transposes on every repetition, which is why this sounds synthetic, whatever went into it.

GRIP
How much comes round again, and LOOP TONE decides what survives a lap - dark below the middle, thin above it. Long feedback needs one or the other, or it only piles up.

The module then shares the stretch's own handles rather than keeping a second set:

WARP RATE
How fast the bend runs, from once every fifty seconds to thirty times a second. Above about ten, it stops being a warp and becomes a sideband.

WARP SHAPE
Sine at one end, triangle in the middle, stepped random at the other: a smooth waver, or a machine losing and regaining its grip.

DRIVE
Saturates the signal on the way in, before the resonance acts on it.

RELEASE
How quickly the gate lets go. Short is a stutter on anything sustained; long holds through the gaps in speech.

LOW
A shelf at 120Hz, fifteen decibels either way.

LO MID
The lower of the two adjustable mids. It is where a mix gets heavy, so it is usually cut rather than lifted.

HI MID AND HIGH (WITH THE TWO MID FREQUENCIES ADJUSTABLE)
The upper mid and the shelf above it, each with its own frequency. The mids move because the instrument does: a preset built around drones and one built around pulses do not put their congestion in the same place.

PRISM

The harmonics family has no stretch. In its place is a prism of twenty-four bands that the incoming sound is played through.

An earlier version analysed the input - a frequency estimated per band by counting zero crossings, an amplitude chased by a follower, a threshold deciding what was present - and drove oscillators from those readings. Every one of those is a guess made forty times a second, and the artefacts were the guesses disagreeing with each other.

BANDPASS and RESONATE are separate levels through the same bands, so either works without the other: a filter at one end, a struck resonator at the other, both at once if wanted. RING is the resonator's decay, and BANDPASS Q is the filter's width.

The bands are placed either geometrically - a third of an octave apart, a spectrum - or harmonically, on whole multiples of the root, where the whole prism is one note, and anything played through it comes back in tune.

LAYOUT
Fades between the two.

STRETCH
Bends the harmonic series, so partial n sits at n to the power of k, which is the difference between a string and a bell.

HARMONISE
Adds a second prism at a just ratio - octave, fifth, fourth, 5/3, 7/4, and their inversions. Just rather than tempered because a 3/2 fifth beats against nothing, whereas seven equal semitones beat against everything, and stacked twenty-four deep that difference is the whole character.

Each band has a fader, labelled by the frequency it sits at, and eight shapes are one press away:

FLAT
Every band open equally. The prism colours nothing and passes what it is given, which is the state to return to after any of the others.

DARK
A spectrum falling away with frequency, which is what most sounds do. The lowest band is open and the highest is nearly shut.

BRIGHT
The same slope the other way up: the top open, the bottom nearly shut. What comes back is air rather than body.

ARCH
One region open and the rest closed - a formant. The peak sits in the middle of the bank and falls away either side, which is how a body with a resonance behaves.

ODD
Odd-numbered bands open, even ones nearly shut. Removing every other partial is what makes a clarinet a clarinet, and it hollows out anything played through it.

THIRDS
Every third band open. Wider than odd and less regular than octaves, so it reads as a chord rather than as a timbre.

OCTAVES
Only the bands nearest octaves of the lowest. The prism stops being a filter and becomes a tuning, and anything played through it comes back in that one chord.

RANDOM
Every band given its own level at random. Pressed twice it is two different instruments, and it is the fastest way to find a shape nobody would have drawn.

After the prism is a reverb whose pitch shift sits inside the feedback loop rather than after it, so a shifted tail shifts again on every pass, and a note climbs away from itself in steps of the interval with thirteen controls:

MIX
The whole filter section's output. The fifty rows below choose which filters are running and how loud each is; this rides all of them together.

DECAY
How long the tail lasts. It stops short of the point where it would never end.

SIZE
How far apart the reflections are, which is heard as the size of the room rather than the length of the tail.

PRE-DELAY
How long the room waits before it answers. A few milliseconds keeps the source distinct from its reflections; longer and the tail arrives as a separate event.

DIFFUSE
How thoroughly the reflections are scattered. Low is a few distinct echoes; high is a wash with nothing countable in it.

DAMP
How much of the top is lost on each pass around the loop, which is what a room full of soft things does.

LOW CUT
How much of the bottom is kept out of the tail. A long reverb on low material turns to mud without this.

SHIMMER
How much of the tail is fed back through the pitch shift. A note then climbs away from itself in steps of the interval instead of arriving once transposed.

INTERVAL
What the shift is: octave down, fifth down, fourth down, unison, fourth, fifth, octave, twelfth. Quantised, because a continuous control here is a detuned smear at almost every setting.

GLIMMER
Brightens the top of the tail at random, so it catches light rather than fading evenly. The difference between a room and a room with something in it.

MOD
Moves the delays slowly. Without it, a long tail settles onto the delay lengths' own frequencies and rings.

WIDTH
The stereo width of everything this chain produces - the stretch, its heads, the loop and the prism where there is one. Mid and side: at nought only what the channels agree on survives, and the whole chain collapses to a point; at 1 it is as the processing placed it, and at 2 their disagreement is amplified until the image is wider than the speakers. It sits last, after the processing rather than before it: setting the width of what goes into a process that decides its own image is close to setting nothing.

YIELD
Pulls the tail down while the input is loud and lets it back when it stops, so a dense passage stays legible, and the reverb is heard in the gaps.

FM MODULE

Two operators became three, and the module became an instrument.

OPERATOR 3
Modulates the modulator rather than the carrier, so its sidebands are sidebands of sidebands - which is where the metallic and bell-like sounds come from.

ALGORITHM
Moves it between that and joining the others on the carrier.

LOCK
Pulls the ratio onto whole numbers. At a whole-number ratio, the sidebands land on harmonics of the carrier and the result is a pitch; anywhere between, they land where they like, and the result is a bell. A continuous knob spends nearly all its travel in the second case, and the first is a handful of points impossible to hit by hand.

BITE, BITE RATE AND BITE DECAY
Put the index under an envelope, since a fixed index is a fixed timbre and a spectrum falling as a note decays is what a struck thing does.

FILTER ENV
Opens the cutoff from the same strike.

CARRIER FB
The carrier fed into its own phase, sine through saw to noise.

FOLD
Harmonics by shape rather than by phase; CUTOFF and RESONANCE after it; DETUNE in cents; DRIFT, a slow wander on the ratio; and GLIDE on the carrier.

Where a control is a frequency or a time, it says so - hertz, seconds, cents, or a multiple - read back through the same expression the synth uses, so the figure on screen is the number the module is using.

MODULATION

Six slots, each holding one modulator, and any slot can be sent to any control at any depth. Nothing is wired in advance: a slot is made, pointed at a fader, and the fader then holds what the modulator says rather than where it was left.

LFO
A shape running at a rate you set, from a cycle a minute to audio speed. Below about ten hertz it is movement; above it, it becomes a tone in whatever it is modulating.

ENVELOPE
One shape, run from a trigger rather than continuously. It is how a modulation happens once per note instead of all the time.

FOLLOWER
Takes its value from how loud something is. What is playing drives what is changing, so the instrument answers itself rather than answering a clock.

SAMPLE-AND-HOLD
A value taken at intervals and held until the next one. It is the difference between a control that slides and one that steps, and stepped modulation is heard as decision rather than as drift.

RANDOM WALK
A value that moves by small amounts from wherever it already is. It never jumps and it never repeats, which is what makes a long passage change without anything appearing to change it.

MACRO
A modulator with no motion of its own: its value is a fader, so one hand movement can drive any number of parameters at any depth, some of them inverted.

Click a control's name to send the selected slot to it; click again to disconnect. A modulated control's own bar shows the value it is actually holding rather than where the fader was left.

CLAIM MAP

Three hundred and ninety-four phrase generators run here, and none of them knows the others exist.

LISTEN
In the field section, gives them one thing in common: a record of what has just been played, one slot per pitch, deposited by every voice and read before the next note. Above zero, each voice moves away from pitches already ringing, and the ensemble spreads across the register; below zero, they converge, and the texture collapses into unisons and octaves nobody wrote. Only the octave is ever changed, so the harmony cannot be damaged.

LISTEN HOLD
How long a pitch stays occupied.

SEQUENCER

Twelve voices on the sequencer's own clock, made to carry weight. Each builds its low end a different way rather than being one method with a tone control, which is why they can be layered without simply getting louder.

SUB
A sine below everything else, tuned to the root and doing nothing but weight. It is felt more than heard, and it is what a mix is missing when it sounds thin at volume.

ROLLER
A repeating low figure on the step rather than a single note per step. It reads as momentum: the same pitch, arriving often enough to become a texture.

WOBBLE
Reads the tail through a moving delay, so it drifts in pitch the way tape does.

STAB
Short, loud and gone. The whole envelope lands inside one step, which is what makes it punctuation rather than material.

DRONE
The harmonic series itself, each partial breathing separately. Nothing is filtered: the movement is the partials disagreeing with each other about how loud they are.

SNARL
Weight with an edge on it - the low end driven until it stops being clean. It occupies the same register as the sub and is heard as attitude rather than as depth.

ARP
A figure played inside a single step. The sequencer gives it one slot and it fills that slot with several notes, so the pattern moves faster than the grid it sits on.

VOWEL
A formant filter over everything, its vowel drifting slowly.

RING
How long the resonators ring, from a knock to twelve seconds. Long enough and a passing sound leaves a chord hanging behind it.

CRUSH
Bit reduction on the master.

FBFM
An operator eating itself: the modulator fed back into its own phase. Low it is a tone, high it is noise, and the whole way between is one control.

STACK
Eight partials added rather than filtered. Building a spectrum up from nothing gives a different weight from cutting one down, and this is the additive half of that pair.

Each makes its weight a different way rather than being a variant of one method - a figure played inside a single step, three formants over a saw, ring modulation, decimation rather than saturation, an operator eating itself, eight partials added rather than filtered.

The kick has a sine beneath it, tuned two octaves below the root and triggered with it, whose pitch falls steeply over the first forty milliseconds. That is what a bold kick is, and why raising the sample level never achieves it.

HOUSEKEEPING

Everything in the studio that nobody is listening to is paused once a second and woken instantly when it is wanted. A synth cannot know it is inaudible - its level is a number on a bus, and reading it does not stop the graph from being computed - so the decision is made from the language.

Every parameter arrives in sixty milliseconds rather than three or four hundred. The smoothing exists to stop a control bus stepping inside a block, which sixty removes as completely as four hundred did; the rest was inertia nobody asked for.

INTERFACE OVERVIEW

The interface is organised into columns that expose the studio's sound sources, global temporal state, processing stages, routing, and system controls. The signal path remains visible throughout, providing a continuous view of how audio flows through the instrument.

The first five stages operate as send effects. Each source determines how much of its signal is sent to each processor, allowing different sound generators to occupy independent processing paths. For example, pulse voices may be gated while sustained voices remain unaffected, or harmonic voices may excite the cavity without affecting the rest of the ensemble.

The final stages are insert processors acting on the complete mix. The Stretch processor and Bus section process the combined output of the instrument rather than individual sources.

The signal chain is displayed at the top of the fourth column. Each stage illuminates whenever audio is routed through it. Module indicators identify the contributing source:

V — Voices
H — Harmonics
P — Pulses
E — External Input
Y — Playback
F — FM

INTERFACE LAYOUT

The instrument is divided into functional columns.

SOURCES
The six families that make sound: the external input, playback, voices, harmonics, pulses and the FM module. Each has its own level and its own sends.

EXTERNAL INPUT
The sound card's input, taken as a voice: its own trim, its own sends, and a lab that no other family can reach.

PLAYBACK
A file read as a voice — on a clock of its own it picks a degree of the scale, transposes the file to it, and moves the head to the place that degree owns.

VOICES
The unpitched and broadly pitched half of the ensemble, and what each one sends on to the sections that treat it.

HARMONICS
The pitched half: drones, bowed metal, tanpura, organ. Each chooses degrees of the current scale rather than frequencies.

PULSES
The sequenced voices and the grid they run on — a clock separate from the field's, with its own tempo, meter, and division.

FM
The modulator: an oscillator pair that is a source rather than a send, so its own level decides whether it is heard at all.

Each source provides an independent level control together with routing to the available processing stages. Level controls affect only the direct signal. A source may remain inaudible while continuing to feed any combination of effects.

STRUCTURE

The global musical state shared by every voice.

SCALE
Which set of intervals every pitched voice chooses from. It is not a key signature but a tuning: changing it changes what the intervals are, not only which notes are used.

ROOT
Base note in MIDI numbers; everything is built up from here.

OCTAVE
Shifts the whole field up or down in octaves.

DENSITY
How often events happen, across every voice at once.

FLOOD
Past what density can do: shorter notes, shorter silences, longer bursts, finer grid.

MICROTONALITY
The scale list is not twelve equal steps and its alternatives. Just, Pythagorean, septimal, undecimal and Bohlen-Pierce tables are here as scales like any other, so a field can be tuned to something no keyboard offers and everything follows it.

TEMPO
How fast the field runs, in beats a minute. It reaches down to one a minute - an event, then a wait long enough to forget it was coming - and up to a hundred and fifty. The sequencer keeps its own clock and is not affected.

DRIFT BEHAVIOUR
Changes made here affect the entire ensemble while preserving the independence of individual voices.

LAB

The Lab provides a shared processing environment containing seventy-two processors.

DYNAMIC FILTERS
Filters whose cutoff is driven by something rather than set - an envelope, a follower, a modulator. The filter moves because the material does.

SIGNAL TRANSFORMATIONS
Processes that change what a sound is rather than how it is balanced: rectification, folding, quantising, ring modulation. None of them is a level or a frequency, and none of them can be undone further down the chain.

REVERBERATION ALGORITHMS
Several kinds of room rather than one with a size control. A plate, a hall and a cloud fail in different ways, and which one is wrong for the material is more useful to know than how large it is.

INFINITE SUSTAINING TEXTURES
Processes with feedback at or past unity, so what enters does not leave. They are the four things in this instrument that keep sounding with every voice at zero.

Processing remains inactive until individual modules are enabled.

EFFECTS

Displays the signal chain in the order encountered by the audio signal and provides access to every processing stage.

SYSTEM

Contains preset management, output routing, metering, recording, and archive functions.

SOUND SOURCES

EXTERNAL INPUT
Processes live audio arriving at the selected audio interface.

Because the level control affects only the direct path, external audio may be routed through any combination of processing stages without being heard dry. This allows, for example, a room microphone to drive the processing network while remaining absent from the final mix.

Four sequencer voices operate directly on a continuously updated four-second recording of the incoming signal, allowing recent audio to be replayed as notes.

PLAYBACK
Plays audio files using either tape-style playback or granular synthesis. With grain set to zero, playback behaves as a tape machine: playback speed determines pitch.

Increasing the grain parameter transitions smoothly into granular playback, allowing time-stretching without pitch change, or transposition without altering duration.

The waveform is displayed in its own editor, where loop points may be adjusted interactively.

Five sequencer voices are dedicated to playback.

FM
A three-operator frequency modulation synthesiser. It is described as two operators further up because that is what it had when the specifications were written; see THE FM MODULE above for the third and what it changed.

Unlike the remaining source modules, the FM engine generates pitched material directly rather than processing existing audio.

DEPTH
Controls the modulation index. At zero, the output is a sine wave; increasing the value progressively enriches the spectrum.

RATIO
Determines the harmonic relationship between carrier and modulator. Integer ratios produce harmonic spectra, while non-integer ratios generate bell-like and inharmonic timbres.

TRACK
Locks the carrier frequency to the global root, allowing the instrument to remain harmonically aligned as the musical field evolves.

FEEDBACK
Routes the modulator into itself, increasing spectral complexity and producing more dynamic, performance-like behaviour.

PROCESSING MODULES

FILTER
A collection of filters organised into 12 banks.

LADDER
The four-pole design a Moog uses, resonance and all. It loses low end as it resonates, which is the part people mean when they call a filter warm.

STATE-VARIABLE
Low, high and band out of one structure, so the same setting can be asked for three different ways. It stays stable where a ladder is already screaming.

COMB
One tuned resonance across the whole field.

ALL-PASS
Changes phase and leaves level alone. On its own it does almost nothing audible; in a feedback loop it is what turns a delay into a room.

RING
How long the resonators ring, from a knock to twelve seconds. Long enough and a passing sound leaves a chord hanging behind it.

SPECTRAL
Filtering done on the analysis rather than on the signal - bins kept, thinned or scrambled instead of frequencies attenuated. It can do things no arrangement of poles can, and it sounds like it.

VINTAGE PEDAL-INSPIRED DESIGNS
Filters modelled on pedals rather than on studio equipment. What they contribute is their faults: a narrow range, an uneven resonance and a character that cannot be dialled out.

Selecting a filter opens its dedicated control panel, allowing multiple filters to be configured simultaneously.

SKEW
A collection of nonlinear wave-shaping processes including:

FUZZ
Up to sixty times gain into a hyperbolic curve, divided back down after, so the level holds while the waveform squares off. Symmetric, and therefore odd in its harmonics.

GERMANIUM
A clipping ceiling that closes from two down to a twentieth as the control rises, with a little of the undistorted signal mixed back in and the result soft-clipped. Softer and less regular than the fuzz.

OCTAVE FUZZ
Full-wave rectification, which folds the negative half of the wave upward and produces a tone an octave above the fundamental. Mixed in rather than driven, so it adds a voice instead of replacing one.

BIT CRUSHING
Quantises the signal, from sixty-five thousand steps down to four. The noise this adds follows the material rather than sitting under it, so it grows and recedes with the sound.

DECIMATION
Holds each sample until the next, lowering the effective rate from eight kilohertz to three hundred hertz. The aliasing it produces descends as the pitch ascends.

WAVEFOLDING
Folds the wave back on itself at a threshold that closes from eight down to under half. It adds harmonics without ever clipping, so the result stays bright rather than square.

TUBE SATURATION
An offset is added before the curve and taken out again after, so the two halves of the wave are shaped by different amounts. That asymmetry is what produces even harmonics rather than odd.

DIODE CLIPPING
The positive half is driven ten times harder than the negative, and each is curved separately, which is the behaviour of a pair of unmatched diodes.

GATED DRIVE
Forty times gain into a hard curve, behind a gate that shuts off anything below a fiftieth of full scale. Loud material is destroyed, and quiet material is silenced.

HARSH DISTORTION
A Chebyshev shaper, which maps the signal through a polynomial rather than a curve, so it produces a fixed harmonic series instead of a saturation. Mixed in.

DRIVE
Saturates the signal on the way in, before the resonance acts on it. Determines the input level entering the processor rather than blending a processed signal with the original.

GATE
How much of the external input goes to the gate. A rhythmic gate with independent rate and chaos controls. Because it operates as a send processor, different source groups may be gated independently.

CAVITY
How much of the input goes to the cavity. A bank of tuned resonant delays with fundamentals ranging from 0.2 Hz upward. Its position after the Skew processor allows harmonic content generated by distortion to excite the resonant network.

LAB

A shared effects environment.

Reverberation
The lab's rooms, from a small bright box to a tail that outlasts the phrase that started it.

DELAY
Three delays at prime ratios, each feeding the next rather than itself. One delay repeats; three at incommensurate times fill in, and the pattern never settles because the ratios never come round.

FILTERING
The lab's own filters, applied to whatever is sent to it rather than to one family. A send is how a single voice gets filtered without the rest of the ensemble following it.

SPECTRAL TRANSFORMATIONS
Freezing, scrambling, thinning and shifting done on the spectrum. The material's colour survives and its behaviour does not, which is what makes a freeze sound like a room rather than like a loop.

PROCESSORS
Seventy-two of them, all inactive until switched on. Nothing here costs anything while it is off, so the list is a menu rather than a chain.

ROOMS
The reverb: how large the space is, how quickly it absorbs its own high end, and how much of it replaces the dry signal.

REVERB
How large the space is, how quickly it absorbs its own high end, and how much of it replaces the dry signal.

PLATES
The dense, bright reverberation of a sheet under tension. No early reflections and no sense of size - it does not put the sound in a room, it puts a sheen on it.

TAPE DELAY
Repeats that lose their top end and their steadiness a little more each pass. The wobble is the point: a perfectly stable delay reads as a copy, and this reads as a machine.

BUCKET-BRIGADE DELAY
An analogue delay that passes the signal along a chain of stages, losing bandwidth at every one. Short settings are a metallic resonance rather than an echo.

PING-PONG DELAY
Repeats alternating between the channels. It makes width out of time, and at short settings it is heard as a position rather than as a repeat.

MULTITAP DELAY
Several taps at different distances off one line. Because they share a line they share a feedback, so a rhythm can be set up and then dissolved by turning one control.

CLOUD REVERBS
Reverberation built from grains rather than from reflections, so the tail can be longer than the sound that caused it and need not resemble it. It is the reverb that does not describe a space.

FREEZE PROCESSING
Holds a moment of the spectrum and keeps sounding it. What is frozen is whatever was passing at that instant, so it is a decision about timing rather than about sound.

DUB ECHOES
Feedback taken to the edge of running away, with the filter inside the loop. Each pass is darker and louder than sense allows, which is the whole idiom.

FILTER MODELS
The lab's copy of the filter section: the same designs available to anything sent here rather than to one family.

TRANSFORMATION ALGORITHMS
The processes that are not filters and not rooms - the ones that return something other than what was sent, and cannot be blended back to neutral.

All processors remain inactive until enabled.

BUS

A stereo bus processor inspired by the SSL bus compressor, extended with feed-forward and feedback blending, side-chain high-pass filtering, three compression ratios, automatic release, and four-band equalisation. The Mix control provides parallel compression. At zero, the processor is bypassed.

SEQUENCER

The sequencer contains voices distributed across three sections:

ROM
One of the sequencer's three sections, holding the voices built from fixed material. What it plays was rendered before the phrase began.

TILDE
The section whose voices are generated as they play rather than read back. It is where the sequencer stops being a player and becomes an instrument.

EXTERNAL
The sound card's input, taken as a voice: its own trim, its own sends, and a lab that no other family can reach. The External section triggers material generated by its corresponding source module.

Each sequencer voice exposes thirty-five independent parameters controlling timing, pitch, routing, filtering, probability, Euclidean generation, ratcheting, rotation, repetition, and playback behaviour. The Build parameter determines which internal synthesis construction is used. When set to Auto, a new build is selected for every note. Selecting 0, 1, or 2 fixes the voice to a specific construction.

ROUTING

Every voice and every bank may be assigned to an individual output or positioned within the stereo field.

Voices assigned explicitly retain their routing independently of subsequent bank changes. Each bank also provides an OFF state, silencing any voice that has not been assigned its own output.

AUDIO INTERFACE

The audio device is selected before the instrument is launched.

Two output channels are allocated automatically during startup, eliminating the need to reconfigure the server after launch.

CONTEXT HELP

Hovering over any control displays a concise description of its function at the bottom of the UI. These contextual descriptions provide immediate reference while working and are intended to replace repeated consultation of the manual.

THE PRISM'S SHAPE

Five controls, on the harmonics only. They are multiplications on levels the prism already computes, so they cost almost nothing and stack in any combination.

TILT
Slopes the whole prism, flat in the middle: at either end the lowest band goes silent while the highest doubles. That shape otherwise had to be drawn across twenty-four faders by hand.

FORMAT, FORMAT WIDTH AND FORMAT AMT
A bell in log-frequency that lifts the bands near it. Every real body has resonances that stay put while the pitch moves, and having them is what makes a fixed spectrum sound like an instrument rather than a chord.

ODD / EVEN
Fades between all partials, the odd ones and the even ones - a clarinet at one end, hollow and organ-like at the other. It changes the family of the sound rather than its shade.

A FILE, PLAYED BY THE STRUCTURE

Six controls at the head of the playback panel. On its own clock, the player picks a degree of the current scale, transposes the file to it, and moves the read head to the place that degree owns.

The scale, the root, and the octave are the instrument's own, so the file speaks in whatever key everything else is in and follows it when that changes. The position is derived from the same degree, so a given note always comes from the same part of the file: play the same phrase twice, and it is the same phrase.

SPREAD
How much of the file the scale covers. Narrow is one region read at different pitches; wide gives each degree its own material, so the file becomes an instrument with twelve different sounds in it.

A bar above the rows draws the mapping: the file end to end, the window SPREAD and FORM cut out of it, a tick per degree where it will send the head, and the head itself.

SECTIONS

Every panel is divided by a rule and a heading: a family reads as direct, sends, stretch, prism, reverb, bands rather than as one column of ninety rows. Clicking a heading folds and unfolds everything between it and the next.

The stretch is four of those - what the stretch is, how the heads are arranged, what is fed back, and what interrupts it. `heads level` and `motion level` were added with them: the first sets how present the extra heads are beside the first, the second scales snap, warp, and jump together.

WHAT IS RUNNING?

Each family carries its own pulse. The meter listens to the voices themselves, rather than to the final output, remaining still when the family has nowhere to go. It comes alive when a family is both sounding and routed, revealing its movement through the system.

Beside each name, a small mark reflects its state: hollow when silent, filled when active, pulsing when sounding. The mark responds immediately to changes in the system; the pulse follows the life of the meter.

Sends behave like pathways through the network. Their labels darken when a connection is open and recede to grey when it is closed. Across a column of forty, the active paths become visible at a glance—a map of where sound is flowing.