Workbench User Guide

A modular sound design environment for building effects chains and polyphonic synthesizer voices.

Overview

Workbench is a real-time audio workstation with two modes of operation:

Effects Mode — Route test signals or live audio through a configurable graph of processing modules. Design and audition effects chains, then export them as recipes.
Synth Mode — Build polyphonic synthesizer voices from oscillators, filters, and effects. Play them via MIDI keyboard or on-screen keys. Shape envelopes, add post-mix effects, and save patches.

Think of it as a virtual Eurorack system: a collection of oscillators, filters, effects, modulators, and analysis tools that you can patch together in any configuration.

Getting Started

First Launch

Select a mode from the transport bar: Effects or Synth
Pick a recipe from the sidebar (bundled presets are included)
Press Play — audio starts immediately

Area	Purpose
Sidebar	Recipe browser — select, create, edit, delete
Detail pane	Inspector view (parameters, metadata) or Graph view (visual patching)
Transport bar	Playback controls, signal source, levels, meters, envelope
Waveform display	Real-time output waveform (green on black, 30 fps)

Switch between Inspector and Graph views using the segmented picker in the transport bar.

Effects Mode

Signal Sources

The signal source menu offers 15 built-in test signals across six categories:

Basic Tones

Sine — Pure fundamental. Good for testing frequency response, distortion characteristics, and filter behavior. The cleanest signal — any harmonics in the output were added by your effects chain.
Sawtooth — Rich harmonic series (all harmonics at 1/n amplitude). The most harmonically complete waveform — excellent for testing filters, since you can hear the cutoff sweep through the full spectrum.
Square — Odd harmonics only (1/n amplitude). Hollow sound. Good for testing how effects respond to sharp transients and DC offset.
Noise — White noise (flat spectral density). The ultimate broadband test signal — reveals resonances, filter shapes, and artifacts that tonal signals might miss.
Silence — No input. Useful for hearing self-oscillating filters, feedback loops, and reverb tails decay to zero.

Tonal Signals

Gated Burst — 200ms tone / 500ms silence cycle. Reveals how effects behave at signal onset and during silence — essential for testing compressor attack/release, noise gates, and reverb tail behavior.
PWM — Pulse width sweeps from 10% to 90% duty cycle. Continuously shifting harmonic content exercises filters and distortion differently than a static waveform.
FM — Two-operator FM with modulating index sweep. Produces complex, time-varying spectra that stress-test any processing chain.

Transients

Impulse — Single-sample click. The broadband transient reveals the impulse response of your entire chain — essentially a snapshot of what the effects "sound like" in one click.
Pluck — Karplus-Strong string model. A natural-sounding decaying tone with rich harmonics. Good for hearing how effects interact with realistic instrument timbres.
Arpeggio — C major arpeggio at 3 notes/sec. Tests how effects handle changing pitch content — particularly useful for pitch-dependent effects like flangers and harmonizers.

Modulated Signals

Sweep — Logarithmic chirp from 80 Hz to 8 kHz over 3 seconds. Methodically exercises the full frequency range — watch the waveform display to see how your chain responds at every frequency.
Sample & Hold — Random pitch jumps at 4 Hz. Abrupt frequency changes reveal artifacts, glitches, and tracking behavior in pitch-dependent processors.

Playback & Live

Sample — Choose from synthesized test patterns (chord, drums, melody) or import your own audio file. The most realistic test — hear how your effects chain sounds on actual music.
Live — Microphone input. Process your voice, guitar, or any sound source in real time. Requires headphones on iOS to prevent feedback.

The Frequency Slider

When using basic tones (sine, sawtooth, square) or tonal signals, the frequency slider lets you sweep from 20 Hz to 2 kHz. Sweep slowly while listening — you'll hear how filters, resonances, and distortion change with input frequency.

Input Level

The input level slider (default 0 dBFS) controls the signal amplitude before it hits your effects chain. This matters because many effects are level-dependent:

Distortion/saturation — higher input = more drive
Compressors — input level relative to threshold determines gain reduction
Reverbs — input level affects how the reverb tail decays

Think of it as the gain knob on a mixer channel before the insert effects.

Building an Effects Chain

Click Edit on a recipe (or New Recipe for a blank slate)
Switch to Graph view
Click Add Node in the toolbar to open the component browser
Search or browse by category — select a component to add it to the canvas
Connect nodes: drag from the output port (right side dot) to the input port (left side dot) of another node
Click any node to select it — the inspector panel shows its parameters
Adjust parameters with sliders while audio plays to hear changes in real time

Signal Flow

Audio flows left to right through connected nodes. When multiple nodes feed into one input, their signals are summed. The final node's output becomes the effect chain's output.

Input → [Node A] → [Node B] → [Node C] → Output
                         ↗
              [Node D] ──┘  (summed into Node C's input)

The Mixer node is special — it crossfades between exactly two inputs using a Balance parameter (0.0 = first input only, 1.0 = second input only). Use it for parallel processing or A/B blending:

Input → [Distortion] ──→ [Mixer] → Output
    └──→ [Clean path] ──↗

Synth Mode

Overview

Synth mode turns Workbench into a polyphonic synthesizer. Each "voice" is a clone of your voice graph — when you play a note, a fresh voice is allocated, its oscillators are tuned to the note frequency, and an amplitude envelope shapes the sound.

Playing Notes

MIDI Keyboard — Connect any MIDI controller. Workbench auto-detects all MIDI sources and shows connection status in the transport bar. Supports note on/off, velocity, pitch bend, and channel pressure.

On-Screen Keyboard — A 2-octave piano keyboard appears below the waveform display. Click or drag across keys to play. Use the octave buttons to shift range (0-7), and the velocity slider to control dynamics.

Voice Architecture

MIDI Note → Voice Graph (per voice)
                ↓
         ADSR Envelope → Stereo Pan
                ↓
            Sum all voices
                ↓
         Post-Mix Effects (shared, runs once)
                ↓
              Output

Each playing note gets its own copy of the voice graph. The voice graph typically contains:

Oscillators — Sound sources, tuned to the note frequency via note bindings
Filters — Shape the timbre
Waveshapers/Distortion — Add harmonics and character

After all voices are mixed to stereo, the optional post-mix effects graph processes the combined output. This is where you put effects that should be shared across all voices — reverb, delay, master EQ, compression. Running these once (instead of per-voice) saves significant CPU and produces a more natural blend.

Note Bindings

Note bindings connect MIDI note data to parameters inside your voice graph. Select a node in the graph editor, and the Note Bindings section appears in the inspector panel.

Source	Description	Typical Target
frequency	Note frequency in Hz (A4 = 440 Hz)	Oscillator Frequency
velocity	0.0–1.0 (from MIDI velocity 0–127)	Filter cutoff, amplitude
gate	1.0 while held, 0.0 on release	Gate-triggered effects
noteNumber	Raw MIDI note (0–127 as float)	Lookup tables, switches
pressure	Channel aftertouch (0.0–1.0)	Filter cutoff, vibrato depth
pitchBend	-1.0 to +1.0 (from pitch wheel)	Oscillator pitch modulation

The most essential binding: frequency → Oscillator Frequency. Without this, all notes play the same pitch.

ADSR Envelope

Every voice has an amplitude envelope that shapes the loudness over the note's lifetime. The four sliders in the transport bar control:

Stage	Range	Description
A (Attack)	1 ms – 2 s	Time to ramp from silence to peak amplitude. Fast (< 10 ms) for percussive sounds, slow (200 ms+) for pads.
D (Decay)	0 – 2 s	Time to fall from peak to sustain level. Creates the initial transient shape. Set to 0 to skip.
S (Sustain)	0 – 100%	Level held while the key is pressed. 100% means no decay. Lower values create plucky, decaying sounds even while held.
R (Release)	1 ms – 3 s	Time to fade to silence after key release. Short for staccato, long for sustained tails.

Amplitude
    ^
peak│    /\
    │   /  \
sus │  /    \________
    │ /              \
    │/                \
    └──────────────────→ Time
      A    D    S    R

Polyphony Controls

Control	Description
Voices	Maximum simultaneous notes (1–16). When exceeded, the oldest voice is stolen.
Glide	Portamento time (0 = instant pitch change). When > 0, pitch slides smoothly between notes.
Spread	Stereo width of the voice field (0% = mono center, 100% = full stereo). Voices are panned across the stereo field using equal-power panning.

Post-Mix Effects

When editing a synth recipe in the graph editor, Add Node presents a choice: Voice Graph or Effects Graph.

Voice Graph nodes run per-voice (N copies running simultaneously). Put oscillators, filters, and per-note processing here.
Effects Graph nodes run once on the summed stereo output. Put reverb, delay, master EQ, and compression here.

Effects graph nodes appear below a "POST-MIX EFFECTS" divider with an orange border, visually separated from voice nodes. They wire together the same way — drag between ports to connect.

Why this matters for CPU: A reverb running in the voice graph costs N x reverb. In the effects graph, it costs 1 x reverb. With 8 voices, that's an 8x savings on your most expensive effects. The sound is also better — one shared reverb blends all voices into a natural space, while 8 independent reverbs sound artificial.

Bundled Synth Recipes

Recipe	Description	Envelope	Post-Mix Effects
Warm Pad	Detuned sawtooth pair through SVF lowpass	Slow attack (300 ms), long release (1 s)	FeedbackNetwork reverb
Simple Lead	Single sawtooth, legato glide	Fast attack (5 ms), medium release (200 ms)	None
Sub Bass	Sine oscillator, mono	Fast attack (5 ms), short release (100 ms)	TapeSaturation (post-mix)

Module Reference

The 69 available modules are organized into five tiers. Each module is a self-contained DSP processor with its own parameters, snapshot visualization, and bypass capability.

Generators

Sound sources that create audio from nothing. In synth mode, bind their frequency parameter to MIDI note frequency.

Oscillator

The primary sound source. Six waveforms:

Waveform	Character	Harmonics
Sine	Pure, clean, subby	Fundamental only
Triangle	Soft, flute-like	Odd harmonics, -12 dB/octave rolloff
Square	Hollow, reedy, woody	Odd harmonics, -6 dB/octave rolloff
Sawtooth	Bright, brassy, full	All harmonics, -6 dB/octave rolloff
White Noise	Hiss, texture, percussion	Flat spectral density
Wavetable	Morphable, evolving	Depends on table + morph position

Wavetable mode loads preset waveform tables and lets you morph between frames using the Morph Position parameter. Multi-frame presets create continuously evolving timbres when modulated. Detune parameter adds pitch offset in cents — use multiple oscillators with small detune values (5–15 cents) for chorus/unison thickness.

Noise Generator

Dedicated noise source with selectable color: - White — Equal energy per frequency. Bright, hissy. - Pink — -3 dB/octave rolloff. Perceptually balanced, natural-sounding. - Brown — -6 dB/octave rolloff. Deep, rumbling, thunder-like.

Karplus-Strong

Physical model of a plucked string. An impulse excites a tuned delay line with feedback filtering — the result is a remarkably realistic plucked string that decays naturally. The damping parameter controls how quickly high harmonics decay (more damping = duller, nylon-like; less = brighter, metallic).

Waveguide

Bidirectional delay line physical model with a Friedlander bow friction model. Simulates bowed strings — the bow pressure and velocity parameters create sustained, violin-like tones that respond dynamically to parameter changes.

Hash Oscillator

Experimental sound source that feeds a phase accumulator through hash functions (MurmurHash, FNV-1a, SipHash, CityHash). Produces deterministic pseudo-random waveforms with a metallic, digital character that changes with frequency. Unique timbres not achievable with traditional synthesis.

Reaction-Diffusion Oscillator

Runs a 1D Gray-Scott reaction-diffusion simulation and reads the result as a wavetable. Produces organic, slowly evolving timbres that respond to the feed/kill rate parameters. Alien, biological-sounding textures.

Stigmergic Oscillator

Simulates ant colony pheromone trails on a 1D grid, reading the pheromone field as a wavetable. Produces swarm-like, emergent textures that self-organize over time.

Filters

Shape the frequency content of a signal. The most fundamental tone-shaping tool.

State Variable Filter (SVF)

The workhorse filter. Chamberlin topology, 2x oversampled for stability at high resonance. Four simultaneously computed outputs:

Type	Effect
Lowpass	Removes highs, keeps lows. The classic synth filter — sweeping cutoff from low to high creates the characteristic "opening" sound.
Highpass	Removes lows, keeps highs. Thins out a signal, removes rumble.
Bandpass	Keeps frequencies near cutoff, removes everything else. Creates a focused, vocal-like quality.
Notch	Removes frequencies near cutoff, keeps everything else. Subtle tonal shaping — sweep it slowly for phaser-like effects.

Resonance boosts frequencies at the cutoff point. At extreme settings (approaching 1.0), the filter self-oscillates — becoming a pure sine wave generator. Use with caution; self-oscillation can be very loud.

Biquad

Precision filter based on the Audio EQ Cookbook. Seven types: lowpass, highpass, bandpass, notch, peaking EQ, low shelf, high shelf. More clinical than the SVF — better for surgical EQ work, less character for creative sound design.

One-Pole

The simplest filter: a single pole, switchable between lowpass and highpass. Gentle 6 dB/octave slope. Useful for subtle high-frequency smoothing or DC removal. Very CPU-efficient.

Dynamics

Control the amplitude envelope and loudness of a signal.

Compressor

Full dynamics processor. Reduces the dynamic range by attenuating signals above a threshold.

Threshold — Level above which compression begins. Lower = more compression.
Ratio — How much to reduce. 4:1 means a signal 4 dB over threshold comes out only 1 dB over. At extreme ratios (20:1+), acts as a limiter.
Attack — How quickly compression engages after signal exceeds threshold. Fast attack catches transients; slow attack lets them through (adds punch).
Release — How quickly compression releases after signal drops below threshold. Too fast causes "pumping"; too slow causes sustained dullness.
Knee — Transition sharpness. Soft knee (high values) applies gradual compression around the threshold for transparent dynamics control.
Makeup Gain — Compensates for volume loss from compression.

Brickwall Limiter

64-sample lookahead peak limiter. Catches peaks before they happen. Always active in the Workbench output chain — your signal passes through one regardless. Use a second one inside your effects chain for creative limiting or to protect downstream effects from clipping.

Transient Shaper

Separates attack and sustain portions of a signal using dual envelope followers, then lets you boost or cut each independently. Boost attack for more punch/snap. Boost sustain for more body. Cut attack for softer, more ambient textures.

Envelope Follower

Tracks the amplitude of the input signal with configurable attack and release times. Useful as a modulation source — the output follows the input's volume contour. Connect it via modulation routing to drive filter cutoff from input dynamics (auto-wah) or to duck other parameters (sidechain-style effects).

Distortion & Saturation

Add harmonics, warmth, grit, or outright destruction.

Overdrive

The most versatile distortion. Seven clipping curves:

Curve	Character
Soft Symmetric	Warm, tube-like, musical. The default — works on everything.
Asymmetric	Even harmonics, transistor-like crunch. Adds 2nd harmonic warmth.
Hard Clip	Digital, aggressive, buzzy. Flat-top clipping at extremes.
Diode	Asymmetric with soft knee — germanium diode character.
Full Rectify	Folds negative signal positive — octave-up effect, buzzy and synthy.
Half Rectify	Removes negative half — extreme harmonics, very aggressive.
Crossover	Zero-crossing distortion — thin, brittle, broken-amp character.

The Topology switch changes whether distortion is in the signal path (standard) or in a negative feedback loop (softer, more compressed response). Pre-Filter shapes the input before clipping — MidBoost pushes mids into distortion harder for a more focused sound.

Tape Saturation

Models the nonlinear behavior of analog tape: asymmetric soft clipping, magnetic hysteresis, and pre-emphasis/de-emphasis EQ curves. Produces warm, musical saturation with even-order harmonics. Subtler and more "analog" than Overdrive — excellent on buses and master output. 2x oversampled to prevent aliasing.

Tube Warmth

Simulates vacuum tube harmonic generation. Adds both even and odd harmonics with independent control over each. Even harmonics (2nd, 4th) add warmth and body. Odd harmonics (3rd, 5th) add edge and presence. 2x oversampled.

Transformer Color

Models the saturation characteristics of an audio transformer: low-frequency saturation, high-frequency rolloff with resonance. Subtle, warming effect — the kind of coloration that makes analog gear sound "expensive." 2x oversampled.

Console Bus

Analog summing bus model: asymmetric saturation, gentle HF rolloff, and a noise floor. The "glue" effect — makes things sound like they went through a real mixing console. Very subtle at low settings.

Wave Shaper

Raw transfer function distortion. Seven shaping curves (tanh, hard clip, sine fold, asymmetric, rectify, crossover) applied directly to the signal. More mathematical and less "musical" than Overdrive — but sine fold produces unique metallic timbres when driven hard.

Soft Clipper

Threshold-based tanh saturation with adjustable knee. Clean below threshold, soft saturation above. The most transparent saturator — use it when you want controlled clipping without changing the character of the sound.

Bit Crush

Reduces bit depth and sample rate for lo-fi digital degradation. Low bit depths produce staircase-like quantization distortion. Low sample rates produce aliasing artifacts. Optional dither reduces quantization harshness at the cost of added noise.

Sag

Power supply voltage sag simulation. Models how tube amplifiers lose headroom under sustained load — the signal compresses and distorts asymmetrically as the virtual power supply droops. Creates a dynamic, responsive distortion that reacts to playing intensity.

Bias Gate

Transistor bias/starve simulation with splutter gating. Starving the bias voltage creates broken, gated, sputtering distortion — like a dying battery in a fuzz pedal. Extreme settings produce rhythmic, choppy artifacts.

Modulation Effects

Time-varying processes that create movement and animation in the sound.

Chorus

LFO-modulated delay (15 ms center) creates pitch detuning that simulates multiple voices playing the same part. Adds width and thickness. Multiple voices with different LFO rates create a lusher, more complex chorus.

Flanger

Short modulated delay (2.5 ms center) with feedback. The comb filtering produces a metallic, jet-like sweep. Higher feedback creates more pronounced resonant peaks. Negative feedback inverts the comb filter character.

Phaser

Cascaded allpass filters (2–12 stages) with LFO sweep. Creates moving notches in the frequency spectrum. Fewer stages = subtle, watery. More stages = deeper, more complex phase cancellation. Classic "swoosh" effect.

Tremolo

Amplitude modulation via an internal LFO. The simplest modulation effect — volume goes up and down rhythmically. Square LFO creates choppy gating; sine creates smooth pulsing; triangle is somewhere between.

Ring Modulator

Multiplies the input by an internal oscillator. Sum-and-difference frequencies create metallic, bell-like, or robotic timbres. At audio rates, produces inharmonic sidebands. At low rates, acts like a tremolo with a different character. Selectable modulator waveform.

Frequency Shifter

Bode frequency shifter via Hilbert transform. Shifts all frequencies by a fixed Hz amount (not a musical interval). Small shifts (1–5 Hz) create a slow, rotating phaser-like effect. Larger shifts produce metallic, inharmonic transformations. Unlike pitch shifting, frequency shifting doesn't preserve harmonic relationships.

Wow and Flutter

Modulated delay that simulates the speed instability of tape machines and turntables. Wow is slow, periodic pitch variation. Flutter is faster, more random. Combined, they create the warm, imperfect character of vintage playback equipment.

Spatial & Reverb

Create the impression of acoustic space, from small rooms to infinite voids.

Feedback Network (FDN Reverb)

The primary reverb. An 8-line feedback delay network with Hadamard mixing matrix, Thiran allpass interpolation, and 4-stage pre-diffusion.

Size — Virtual room dimensions. Small values = tight, boxy. Large values = halls, cathedrals.
Feedback — Decay time. Higher values = longer tails. At extreme settings, reverb sustains indefinitely (infinite reverb).
Damping — High-frequency decay rate. Higher = darker reverb that loses brightness over time (like a room with soft surfaces). Lower = brighter, more metallic.
Mod Rate / Depth — Internal LFO modulates delay lengths to reduce metallic coloration. Subtle modulation (< 1 Hz, low depth) adds natural movement.
Diffusion — Pre-diffusion smooths early reflections. Higher = smoother, more "washed" sound. Lower = more discrete echoes.
Mix — Dry/wet blend.

Delay Line

Circular buffer delay with feedback. The fundamental echo effect.

Delay Time — 0–2000 ms. Short times (< 30 ms) create comb filtering and metallic tones. Medium (50–200 ms) creates slapback echo. Long (> 300 ms) creates rhythmic repeats.
Feedback — Echo regeneration. Higher = more repeats. Above 1.0, echoes grow louder (use with limiter!). Soft clipping on the feedback path prevents runaway.
Interpolation — Hermite cubic (default) produces cleaner modulated delays than linear. Matters when automating delay time.

Allpass Filter

Passes all frequencies at equal amplitude but shifts their phase. On its own, nearly inaudible on steady-state signals. In combination (cascaded or with feedback), creates the basis for reverbs and phasers. The delay-based variant creates diffusion effects.

Pitch & Granular

Transform pitch and time-domain characteristics.

Pitch Shifter

Granular time-domain pitch shifting with Hann-windowed grains. Shifts pitch up or down while maintaining duration. Small shifts (< 1 semitone) create subtle detuning. Large shifts create harmonizer effects. Quality degrades at extreme shifts — this is time-domain, not spectral.

Granular Engine

Slices incoming audio into tiny grains (1–500 ms) and reassembles them. The grain parameters create textures ranging from subtle time-stretching to complete sonic destruction:

Grain Size — Small (< 20 ms) = buzzy, granular texture. Medium (50–100 ms) = recognizable but transformed. Large (> 200 ms) = smooth, time-stretched.
Density — Grains per second. Low density = sparse, pointillistic. High density = dense, cloud-like.
Pitch — Grain playback speed/pitch. Can transpose without changing speed.
Position / Scan Rate — Where in the buffer grains are read from. Fixed position = frozen, stuttering. Scanning = time-stretched playback.
Scatter — Random offset added to grain position. Breaks up regularity, adds organic feel.

Spectral Freeze

FFT-based magnitude freeze with advancing phases. Captures a spectral snapshot and sustains it indefinitely — like an infinite reverb that perfectly preserves the tonal content of one moment. The "freeze" effect heard in ambient music. Mix parameter blends between frozen and live signal.

Modulation Sources

These modules generate control signals rather than audio. Connect them to parameters on other modules using modulation routing.

Chaos Attractor

Four strange attractors (Lorenz, Rossler, Henon, Thomas) generate complex, non-repeating modulation patterns. The output is a 3D trajectory — x, y, and z can each modulate different parameters for correlated but non-identical movement. Rate controls how quickly the system evolves. Produces organic, evolving modulation that never exactly repeats.

Step Sequencer

16-step pattern generator. Each step holds a value, and the sequencer advances through them at a configurable rate. Four direction modes (forward, reverse, ping-pong, random). Glide smooths transitions between steps for portamento-like modulation. 15 built-in pattern presets cover pitch sequences, gate patterns, and contour shapes.

Stochastic Source

Eight stochastic processes for random modulation:

Process	Character
Brownian	Drunken walk — smooth, wandering, unpredictable
Ornstein-Uhlenbeck	Mean-reverting random — wanders but returns to center
Perlin	Smooth coherent noise — organic, flowing
Levy	Mostly calm with occasional large jumps — volatile, spiky
Poisson	Random impulses at average rate — sparse, trigger-like
Shot Noise	Decaying impulse bursts — crackly, textured
Markov	State transitions — switches between levels with memory
Sample & Hold	Stepped random — classic synth randomness

Knot Modulator

Traces 3D mathematical knots (Trefoil, Figure-Eight, Torus, Cinquefoil, Lissajous) as modulation sources. Each knot shape produces a unique pattern of x/y/z modulation with different symmetries and periodicities. More structured than chaos, more complex than simple LFOs.

Avalanche Modulator

Hash cascade modulation with extreme sensitivity to input. Small parameter changes produce large output shifts — the "butterfly effect" as a modulation source. Produces crackling, volatile modulation patterns.

Slew Limiter

Rate limiter with independent rise and fall times. Smooths abrupt parameter changes — fast rise + slow fall creates envelope-follower-like behavior. Useful for taming spiky modulation sources or adding portamento to stepped modulators.

Analysis

Passive modules that extract information from audio without modifying it.

Pitch Tracker

YIN-based pitch detection. Identifies the fundamental frequency of monophonic input and outputs it as a snapshot value. Shows detected note name, frequency, cents deviation, and confidence. Useful for monitoring or as a modulation source (pitch-following effects).

Onset Detector

Spectral flux transient detection. Identifies note onsets, percussive hits, and other transient events in the input. Optional band-split mode separates detection across frequency bands. Useful for triggering effects or modulation on beat.

Exotic Processors

Unusual, experimental modules that use unconventional algorithms.

Moebius Folder

Double-length delay with mathematical "twist" operations. Creates strange, folded delay effects where the signal seems to fold back on itself. Produces sounds ranging from subtle thickening to disorienting spatial effects.

Cipher Stream / Cipher Block / Cipher Duet

Apply cryptographic transformations (reduced-round ChaCha, AES-inspired, etc.) to quantized audio. The encryption/decryption process introduces structured but unpredictable distortion. At low bit depths, produces rhythmic digital artifacts. At higher depths, creates subtle metallic textures.

Reaction-Diffusion Filter

Applies Gray-Scott reaction-diffusion dynamics to spectral bins. Produces organic, self-organizing spectral transformations — frequencies grow, spread, and interact like chemical patterns.

Hydraulic / Wind / Thermal Erosion

Three erosion algorithms applied to spectral magnitudes. Each simulates a different natural erosion process (water, wind, heat) on the frequency spectrum, gradually wearing away or redistributing spectral energy. Subtle, evolving timbral transformations.

Immune Processor

Genetic algorithm that evolves a population of "antibodies" to discriminate spectral features. The system learns to recognize and react to specific frequency patterns in the input. Produces adaptive, evolving filtering effects.

Mycorrhizal Network

Graph-based voice network with Hebbian learning. Multiple processing nodes connected in a fungal-network topology learn from the input signal and adapt their connections over time. Self-organizing, emergent processing.

Allergic Reactor

Monitors spectral band energies and triggers reactions when thresholds are exceeded — like an immune system that reacts to specific frequency content. Creates adaptive, responsive effects that change behavior based on what they hear.

Origami Fold

Spectral fold operations (Valley, Mountain, Squash, Petal) applied to frequency-domain data. Each fold type transforms the spectrum in a geometrically-inspired way, producing unique timbral shifts.

Modulation Routing

Modulation brings patches to life by creating movement and interaction between modules. Any modulation source (Chaos Attractor, Step Sequencer, Stochastic Source, etc.) can modulate any parameter on any other module.

Creating a Modulation Route

In the graph editor toolbar, click Add Modulation
Select a source module (must be a modulation source)
Select a target module and parameter
Set depth — how much the modulation affects the parameter
Choose a curve and polarity

Curves

Curve	Shape	Use
Linear	Straight line	Even, proportional response
Exponential	Slow start, fast end	Subtle at low modulation, dramatic at high
Logarithmic	Fast start, slow end	Quick response with gradual saturation
S-Curve	Smoothstep	Gentle transitions at extremes, quick through middle

Polarity

Polarity	Range	Use
Bipolar	-1 to +1	Modulation swings above and below the base value
Unipolar	0 to +1	Modulation only adds to the base value

Multi-Output Sources

Chaos Attractor and Knot Modulator output three simultaneous values (x, y, z). Each can be routed independently to different parameters, creating correlated but distinct modulation on multiple targets.

Transport Controls Reference

Common Controls (Both Modes)

Control	Description
Play / Stop	Start or stop audio processing
Input	Input level in dB (before processing)
Limit	Brickwall limiter ceiling in dB
Bypass	Skip all processing — hear raw source through limiter
Auto Level	RMS-based auto-gain normalization (-18 dBFS target)
Profile	Enable per-node CPU timing (adds overhead)
Inspector	Toggle node inspector in detail view
CPU meter	Real-time CPU load, peak, and glitch count

Synth Mode Controls

Control	Description
Voices	Maximum polyphony (1–16)
Glide	Portamento time (0 = instant)
Spread	Stereo voice width (0–100%)
A / D / S / R	ADSR envelope (attack, decay, sustain, release)

Recipes & Patches

Saving

In the graph editor toolbar, click Save to write the current recipe to the user directory. Saved recipes appear in the sidebar with a user badge and can be reverted to the bundled version if one exists.

Exporting

Export (effects mode only) writes the recipe as a standalone JSON file that can be shared or imported into other Touchstone-based applications.

Recipe Structure

A recipe is a JSON file containing: - Graph — Nodes (modules) and edges (connections) - Exposed Parameters — Parameters surfaced to the user (starred in the inspector) - Modulation Routes — Source → target parameter connections - Metadata — Name, description, category

A synth recipe additionally contains: - Note Bindings — MIDI note properties → node parameters - Polyphony Config — Voice count, steal mode, glide, ADSR - Effects Graph — Optional post-mix effects chain

Tips & Techniques

Building a Subtractive Synth Voice

Start with a harmonically rich oscillator (sawtooth or square)
Bind frequency to the oscillator
Add an SVF in lowpass mode after it
Set filter cutoff around 2–4 kHz, resonance 0.2–0.4
Expose filter cutoff as a parameter for real-time control
Add a post-mix reverb in the effects graph
Set attack to 5–50 ms, release to 200–500 ms

Creating Unison/Detune Thickness

Add multiple oscillators with the same frequency binding but different Detune values (try 5, -5, 10, -10 cents). Mix them with a Mixer node. This is more CPU-efficient than increasing the voice count — 3 oscillators inside 1 voice is cheaper than 3 voices with 1 oscillator each.

Parallel Processing

Use a Mixer node to blend a processed and unprocessed path:

Input → [Heavy Distortion] → [Mixer] → Output
    └─────────────────────────↗

Sweep the Balance parameter to blend between clean and destroyed.

Self-Oscillating Filter Drone

Set signal source to Silence
Add an SVF with resonance at 0.95+
The filter will self-oscillate at the cutoff frequency
Modulate cutoff with a Chaos Attractor or Step Sequencer
Add reverb after the filter

Feedback Patch

Connect a Delay Line's output back to its own input (or through other effects back to the delay). The feedback creates evolving, self-generating textures. Keep the limiter engaged — feedback loops can get loud.

Granular Freeze

Set Granular Engine density high, grain size medium (50–100 ms)
Set scan rate to 0
Position controls which moment is frozen
Modulate position slowly with a Stochastic Source (Perlin) for drifting, textural pads

Workbench User Guide

Overview

Getting Started

First Launch

Navigation

Effects Mode

Signal Sources

Basic Tones

Tonal Signals

Transients

Modulated Signals

Playback & Live

The Frequency Slider

Input Level

Building an Effects Chain

Signal Flow

Synth Mode

Overview

Playing Notes

Voice Architecture

Note Bindings

ADSR Envelope

Polyphony Controls

Post-Mix Effects

Bundled Synth Recipes

Module Reference

Generators

Oscillator

Noise Generator

Karplus-Strong

Waveguide

Hash Oscillator

Reaction-Diffusion Oscillator

Stigmergic Oscillator

Filters

State Variable Filter (SVF)

Biquad

One-Pole

Dynamics

Compressor

Brickwall Limiter

Transient Shaper

Envelope Follower

Distortion & Saturation

Overdrive

Tape Saturation

Tube Warmth

Transformer Color

Console Bus

Wave Shaper

Soft Clipper

Bit Crush

Sag

Bias Gate

Modulation Effects

Chorus

Flanger

Phaser

Tremolo

Ring Modulator

Frequency Shifter

Wow and Flutter

Spatial & Reverb

Feedback Network (FDN Reverb)

Delay Line

Allpass Filter

Pitch & Granular

Pitch Shifter

Granular Engine

Spectral Freeze

Modulation Sources

Chaos Attractor

Step Sequencer

Stochastic Source

Knot Modulator

Avalanche Modulator

Slew Limiter

Analysis

Pitch Tracker

Onset Detector