// ZPWR-SYNTH — DOCS

Overview

The whole synth is one modular patch graph. The graph engine is the shared zpwr-patch-core (libs/zpwr-patch-core, a git submodule) — the same signal-agnostic cable-routing core that powers zpwr-fx. A patch — a grid of DSP nodes wired by per-input source selection — is the voice, instantiated across a polyphonic pool.

It is a fully modular instrument in the lineage of VCV Rack, Cherry Audio Voltage Modular and NI Reaktor Blocks: you build the entire voice by wiring DSP blocks with cables.

Part of the MenkeTechnologies audio stack alongside zpwr-fx, zpwr-midi-fx, and zpwr-patch-core.

Formats & Platforms

Builds as AU, VST3, CLAP and Standalone on macOS (Apple Silicon / Intel) and Linux (x86_64 / aarch64). On macOS the default build is a universal binary (x86_64;arm64), so the VST3/AU/CLAP load on both Intel and Apple Silicon hosts. Windows builds VST3 + CLAP (AU is macOS-only).

The Standalone hosts the full WebView editor and takes MIDI from any connected input. User presets are .zsynthpatch files under ~/Library/zpwr-synth/Presets.

Patch Model

The default voice is Osc(Note) → VCA ← Env(Gate). There is no fixed signal path: a patch — a grid of DSP nodes wired by per-input source selection — is the voice, instantiated across a polyphonic pool. Each node carries a type, 4 inputs, up to 12 params and one output:

ZpwrSynthProcessor (JUCE AudioProcessor)
  └─ zsynth::PolyEngine                 # pool of voices, one zpc::RuntimeGraph each
       └─ zpc::RuntimeGraph (per voice) # the shared core graph, evaluated per sample
            ├─ nodes[N]   each: type + 4 inputs + up to 12 params + 1 output
            ├─ external sources fed per voice (id → value, supplied each sample):
            │    In L/R, Noise, Soft Keys (expandable pool), Note, Gate, Velocity, Mod Wheel,
            │    Pitch Bend, Aftertouch, MPE Pressure / Slide / Bend
            ├─ mods[ ]     route list: source → (node, param) × depth
            └─ out L / out R  (each selects a source)

The synth-specific node types (dsp/SynthModules.cpp, namespace zsynth) are a DSP pack registered onto the core registry; the generators (Osc, Wt, Sample, Granular) read each voice's Note/Gate straight from the core's ComputeContext::external array, so one patch definition plays polyphonically. The patch is saved as JSON in plugin state. MIDI/MPE reaches the synth from the host or the standalone's MIDI input.

Modules

The full catalog — every patchable block with its inputs and parameters — lives in the Module Reference, generated from the live registry. The synth-specific generator and shaper nodes added on top of the shared core include:

• Osc — analog oscillator (note-driven): wave, octave, fine, PW
• Wt — wavetable oscillator: table, position, octave, fine
• Supersaw — 7 detuned saws (JP-8000 style): octave, detune, mix
• FM — 2-operator phase-modulation sine: ratio, index, octave
• Karplus — plucked-string physical model: octave, damping, feedback
• Additive — summed sine harmonics: partials, rolloff, octave, odd/even
• Sync — hard-sync sawtooth: tune, sync ratio, octave
• ChordOsc — one note → 3-note chord of saws: type, octave, detune
• HardKick — hardstyle/rawstyle kick (gate-struck pitch-sweep + tanh drive): tune, punch, psweep, decay, click, drive
• Screech — hardstyle/gabber screech lead (detuned saws → drive → formant BP): octave, fine, detune, drive, formant, reso
• Hoover — Alpha Juno hoover / Mentasm: octave, fine, PWM, chorus, sweep, sweepT
• Reese — DnB / neurofunk Reese bass (detuned beating saws → LP → drive): octave, fine, detune, voices, tone, drive
• Sub — sub-oscillator below the note: octave, wave, level
• Noise — white/pink noise source: color, level
• Sample / Granular — sample playback / granular: slot, start/pos, size, rate
• Env — ADSR (gate-driven): A, D, S, R
• VCA — in1 × in2 (audio × CV)
• Filter — TPT state-variable: cutoff, reso, mode, mod
• Folder — sine wavefolder (west-coast): fold, bias, mix
• Waveshaper — 4-curve shaper: drive, shape, mix
• Crusher — bit + sample-rate reduction: bits, downsample, mix
• Glide — portamento on the note CV: time
• SampleHold — clocked sample-and-hold CV source: rate, glide
• Vector — 4-source XY-morph oscillator (Prophet-VS): octave, X, Y, detune
• DiodeLadder — 4-pole diode-ladder LP (TB-303 grit): cutoff, reso, mod, drive
• StepLFO — stepped LFO (stair / random S&H): rate, steps, shape, smooth
• NoiseLFO — coloured-noise CV (white/pink/brown) + S&H: rate, color, depth
• Scaler — scale/offset/curve a CV (waveshaper for mod): scale, offset, curve
• Delay / LFO / RingMod / Drive / Gain / Mixer — shaping + modulation, per-type

Every oscillator (Osc, Wt, Supersaw, FM, Sub, Sync, Additive) has a Voices (1–11) unison param plus Detune (cents) — detuned copies summed and loudness-normalised. Voices = 1 is the classic mono behaviour, so existing patches are unchanged. A first-class Trigger modulation source (a 1-sample impulse on each note-on edge) sits alongside Note / Gate / Velocity for modular-style routing.

Modulation

Modulation is the patching itself: any LFO / Env node output or soft key patched into a node input — or mod-routed to a param — is a mod-matrix connection. The editor exposes a per-param mod matrix; the per-voice graph carries a route list (source → (node, param) × depth).

Per-voice external mod sources fed to every patch each sample are Note, Gate, Velocity, Mod Wheel, Pitch Bend, Aftertouch, MPE Pressure / Slide / Bend, Noise, In L/R, and the soft keys (an expandable pool). A first-class Trigger source (a 1-sample impulse on each note-on edge) sits alongside Note / Gate / Velocity for modular-style routing.

Soft Keys & Automation

Host-automatable parameters are the soft keys (sk0…, an expandable pool, ceiling 32; 16 active by default, the active count saved in plugin state), master_in / master_out / master_bypass, plus a reserved pool of 96 Auto N host params — right-click any block param to bind it (the Kontakt/Reaktor model), so any module parameter becomes host-automatable / CC-mappable. 32 soft keys + 96 auto params = 128 CC-mappable slots, one per MIDI CC. Everything else lives in the patch (saved as JSON in plugin state).

User Interface

The editor is a WebView hosting the shared zpwr-patch-core cyberpunk patcher (libs/zpwr-patch-core/webui, served from BinaryData) — byte-for-byte the same UI zpwr-fx uses: drag-to-patch neon cables (fan-out, feedback, per-cable gain + colour), a dynamic block grid (+ ADD BLOCK / delete any number, no fixed node count), double-click a block for its detail modal, and a per-param mod matrix. Panes:

• PATCH — the node grid + cable routing + an expandable row of soft-key macro knobs (+/− to add/remove; 16 active by default). The toolbar carries an INIT button (unplug every cable & mod, keep the blocks) and 🗑 (blank the whole patch). An ⚡ EZ MODE toggle lays down (and keeps) a playable voice — generators(Note) → VCA ← amp Env(Gate), then VCA → Filter (cutoff swept by a second filter Env) → out. While on, oscillators you add are summed straight into the voice; your generator types are kept. Two toolbar toggles drive the stereo image: Stereo mirrors every block into a right-channel clone (an independent dual-mono voice per side), and Stereo Lock (shown only when Stereo is on) keeps the two channels in sync — moving a knob on either side moves its clone by the same delta, and the mirror is bidirectional.
• SETTINGS — master in/out (dB) + bypass, Auto Gain Stage + target, the brickwall limiter, the rest of the audio-engine settings, about.

Auto gain staging

Two Settings (both on by default) keep the signal between blocks from clipping no matter how hot the patch-cable gains are, per block, inside every voice graph (and the FX buses):

• Auto Gain Stage rides levels — a fast-attack / slow-release peak follower whose smoothed gain (≤ 1) pulls each block toward the Auto-Gain Target ceiling.
• Soft Clip is the guarantee — an instant (sample-accurate) tanh bound at the same ceiling on every block output, catching the fast transients the AGC's ≈2 ms attack would miss.

The live cable glow tracks each block's level; turn either off in Settings for raw gain. Distinct from the master Brickwall Limiter (a single hard ceiling on the final output).

Perform

The PERFORM pane is a play surface with no patching. A PRESET MORPH pad bilinearly interpolates between four corner presets (A/B/C/D, host-automatable morphX/morphY so it runs editor-closed; 🎲 fills all four corners at random). An Omnisphere-style ORB: drag the puck where the angle selects one of 8 randomised scenes and the distance from centre scales intensity, 🎲 rolls fresh scenes, and ⏺ / ▶ record the orb gesture and loop it back. XY macro pads (each drives a pair of soft keys, per-pad HOLD/SPRING release), a row of macro knobs, eight macro-surface SNAPSHOTS (click empty to save, filled to recall, right-click to clear), dice/🎲 RANDOMIZE for all macros, scale/key quantize + CHORD stacking (Oct/5th/Maj/Min/Maj7/Min7/Sus4/Power), and an on-screen keyboard with pitch-bend / mod wheels. The control panel also carries the MIDI IN toggles (PROGRAM = respond to MIDI Program Change, BANK = respond to Bank Select CC0/CC32; both default ON) and the ARP controls — mode (Up/Down/Up-Down/Random/As-Played), rate (1/4…1/16T) and LATCH.

The editor also has an on-screen keyboard at the bottom plus computer-keyboard input: the home/QWERTY rows (A W S E D F T G Y H U J K …) play a chromatic octave around C4. Click a key once to give the keyboard focus, then type.

Presets

The PRESETS pane ships 256 general factory voices across Factory 1 + Factory 2 (128 each; category-prefixed names: BA bass, LD lead, PD pad, KY keys, PL pluck, BR brass, BE bell, ST strings, DR perc, SEQ/FX etc.) spanning subtractive, FM, additive, supersaw, wavetable, vector, sync and Karplus voices, plus three genre banks designed from documented production techniques — an uplifting Trance bank, a Hard Techno bank, and a Schranz bank (≈160 BPM). Each is generated by an offline tool in tools/gen_*_bank.cpp (shared scaffolding in tools/preset_gen.h). Every preset carries facet tags (Type / Character / Style / Author / Desc) stored in the preset JSON, so the browser's TYPE / CHARACTER / STYLE / BANK facets are fully populated. Plus user presets (.zsynthpatch under ~/Library/zpwr-synth/Presets).

NKS export — every factory voice can be written as a Native Kontrol Standard .nksf preset (RIFF/NIKS container: NISI summary metadata from the facet tags, PLID plugin match, PCHK = the real plugin state, NICA controller page). Each also gets a Komplete Kontrol preview — a 7 s offline render (single sustained C3 + release tail, faded, 44.1 kHz Ogg Vorbis) written to <bank>/.previews/<name>.nksf.ogg. Launch the standalone with ZPWR_EXPORT_NKS=<dir> to export every factory voice into one subfolder per bank.