Tract Plugin Pack

May 31, 2026 · View on GitHub

A collection of audio effect plugins (VST3, CLAP, standalone) built with nih-plug in Rust.

Plugins

Gain Brain

A lightweight gain utility with cross-instance group linking. Multiple instances can be assigned to the same group (1-16), and changing gain on any grouped instance applies that change to all others.

  • 16 groups with Absolute (identical values) and Relative (delta-based) link modes
  • Invert toggle for mirrored gain movement (ducking, sidechain-style workflows)
  • In-process shared state with cumulative canonical delta sync (lock-free atomics, zero overhead)
  • CPU-rendered GUI with rotary dial (tiny-skia + softbuffer), freely resizable
  • ~0.62 MB RSS and 0.03% CPU per instance (200 instances grouped @ 6.3% CPU, 123 MB total)
  • Inspired by BlueCat's Gain Suite

GS Meter

A lightweight loudness meter with integrated gain utility, purpose-built for clip-to-zero workflows. Designed to run 100+ instances per project without significant CPU or memory impact.

  • dB and LUFS modes with per-mode gain/reference and gain-match buttons
  • Peak max, true peak (ITU-R BS.1770-4), RMS, EBU R128 (integrated, short-term, momentary, LRA)
  • CPU-rendered GUI (tiny-skia + softbuffer), freely resizable -- no GPU drivers loaded
  • SIMD-optimized metering (f32x16)
  • ~1.8 MB RSS and 0.05% CPU per instance (300 instances @ 15% CPU, 560 MB total)
  • Inspired by TBProAudio dpMeter

HD26

A unison-detune chorus and pseudo-stereo widener, inspired by Serum's Hyper/Dimension module. Two sections run in series, labelled H____ and D________ on the panel. The first stacks up to 7 modulated fractional-delay voices per channel for rich detuned unison; the second is a 4-tap pseudo-stereo widener with two modes — AM (antisymmetric Side injection, mono-safe) and Pitch (delay modulation with cross-feed). A transient-detector Retrig re-aligns voice phases on detected attacks.

  • Unison chorus (H____): 0–7 voices (0 = off), Szabo-style asymmetric detune spread, per-voice modulated fractional delay (Catmull-Rom interp), detune/rate/width controls; a depth clamp keeps the modulation sweep within the delay line's range
  • Widener (D________): 4-tap widener in AM mode (antisymmetric L/R side injection → mono sums to exact dry) or Pitch mode (modulated delay + cross-feed); a wet HPF keeps the low end out of the widened path
  • Retrig: fast/slow-envelope onset detector with refractory period; fires when a transient is detected to reset voice phases (reduces smear on new notes)
  • Zero latency; no FFT; no allocations on the audio thread
  • CPU-rendered GUI with lock-free retrig LED and level-bar telemetry (tiny-skia + softbuffer), freely resizable

Imagine

A multiband stereo imager modeled on iZotope Ozone Imager. Four fixed bands with switchable linear-phase FIR or Linkwitz-Riley IIR (Lipshitz/Vanderkooy compensated) crossovers. Per-band Ozone-style Width (S_gain = (width+100)/100, mid unchanged), two Stereoize modes — Mode I is a Haas mid-into-side delay (1–20 ms control), Mode II is a Schroeder/Gerzon all-pass decorrelator (0.5–2.0× delay-scale control) — and a global Recover Sides that folds a Hilbert-rotated residue of removed-side energy back into mid for perceptual width retention when narrowing.

  • 4 fixed bands with 3 draggable crossover splits, log-frequency display
  • Ozone-style Width law: scales the side channel from 0 (mono-fold) through 1 (unity) to 2× (max widening), mid unchanged
  • Stereoize per band with on/off toggle — Mode I (Haas, 1–20 ms tap) or Mode II (real Schroeder/Gerzon decorrelator, 0.5–2.0× delay scale, xcorr < 0.3 on broadband noise)
  • Global Recover Sides for narrowing-without-energy-loss workflows
  • Four vectorscope modes — half-disc Polar dot cloud, Polar Level rays, Goniometer (45°-rotated dual-tone dots), traditional Lissajous (X=L, Y=R) — with correlation + balance bars and a per-channel L/R peak meter strip in the half-disc modes
  • Spectrum + magnitude-squared coherence γ²(k) display via single complex M+jS FFT
  • CPU-rendered GUI (tiny-skia + softbuffer), freely resizable, Cassiopeia A gold/teal palette
  • Inspired by iZotope Ozone Imager

Miff

A convolution filter whose FIR kernel you draw by hand with an MSEG editor — a sibling of Wavetable Filter, but the kernel comes from a sketched curve instead of a wavetable file.

  • The drawn curve is the FIR impulse response; a bipolar tap map (2·value − 1) places the editor's 0.5 midline at a zero tap, enabling highpass/bandpass/comb shapes by drawing above or below centre
  • Raw Mode: direct time-domain SIMD convolution (zero latency)
  • Phaseless Mode: fixed 4096-point STFT magnitude-only filtering (constant 2048-sample latency)
  • Adjustable kernel Length (64–4096 taps); peak-magnitude normalization keeps loudness consistent and never boosts above 0 dB
  • MSEG editor with grid snap, Alt-drag stepped-draw, and a styled randomizer; frequency-response view with live input-spectrum shadow
  • A flat default curve bakes to an all-zero kernel — a fresh instance is clean dry passthrough
  • CPU-rendered GUI (tiny-skia + softbuffer), freely resizable

Multosis

A 16-row grid sequencer where each row runs a different audio effect driven by per-row MSEG modulators. Effects process serially — row 1's output feeds row 2 and so on — so the grid composes a programmable chain whose parameters sweep in time with the host.

  • 50 registry entries (a None passthrough plus 49 effects) across 9 families: Filter (SVF, Comb, Ladder, Diode, Sallen-Key, Phaser Filter), Distortion (Bitcrush, Distortion, Downsample, Wavefolder), Dynamics (Compressor, Gate, Limiter, Transient Shaper), Modulation (Auto Pan, Chorus, Flanger, FM, Phaser, Ring, Tremolo, Vibrato), Pitch (Frequency Shift, Pitch Shift, Varispeed), Spatial (Haas, Stereo Widener), Time (Delay, Plate, Repeat, Reverb, Stretch), Misc (Vocoder), and a 16-effect Spectral family — 14 effects sharing a switchable-FFT-size spectral engine, plus the fixed-FFT Satch and Warp Zone
  • 16 rows × 32 columns, with a draggable loop region and free / beat-synced playback
  • 4 MSEGs per row (3 assignable + always-on Amp) with Bipolar/Unipolar polarity, Beat-synced or Time-synced loop length, Snap-to-grid with Shift-to-bypass, six randomizer styles, and right-click Transform menu for compress/expand on values or times
  • Trigger sources include free Hz, beat divisions, and audio-driven transient detection
  • Wet-bus compressor between the chain output and the global dry/wet blend
  • Per-row mute/solo and a PDC (plugin delay compensation) stripe that lights only on active rows carrying latency
  • CPU-rendered GUI (tiny-skia + softbuffer), freely resizable

Nap

A character reverb built on the Extended Dark Velvet Noise (EDVN) engine. The defining feature is that you draw three curves — Decay, Width, and Tone — over a shared tail-position axis, sculpting how the tail's loudness, stereo spread, and tonal colour each evolve along their length. You can draw exponential decays, gated shapes, reverse swells, plateus, or anything in between; the Width curve collapses the image to mono or fans it wide at any point in the tail; the Tone curve scripts a bright-to-dark air-absorption arc or any tonal trajectory.

  • Three independent MSEG curve editors (Decay / Width / Tone) sharing a 0 – 100 % tail-position axis
  • Dual convolution engine — Zero Latency (time-domain O(M) tapped-delay sparse convolution, no FFT, zero reported latency) and Efficient (UPOLS FFT convolution of the analytically baked IR, ~512-sample latency, lower CPU at large/dense settings; both engines produce identical output)
  • Live tail visualization in the Decay pane shows the generated pulse field (height = gain, colour = tone/filter, L/R split = width jitter)
  • Width curve drives per-pulse right-channel jitter for frequency-dependent interaural coherence (BDVN coherence-via-jitter theory)
  • Design-time controls (Size, Density, Width amount, Seed) re-generate the velvet sequence off the audio thread; automatable controls (Mix, Pre-Delay, Input, Output) smooth continuously
  • CPU-rendered GUI (tiny-skia + softbuffer), freely resizable

Pope Scope

A multichannel real-time oscilloscope with beat sync. Multiple instances share audio data through a global store, allowing one window to display waveforms from up to 16 tracks simultaneously.

  • Three display modes: Vertical (stacked), Overlay (superimposed), Sum (mixed)
  • Three draw styles: Line, Filled, Both
  • Beat sync mode with bar/beat grid alignment (1/4, 1/2, 1, 2, 4 bars)
  • Free-running mode with adjustable timebase (1ms - 10s)
  • dB-scaled amplitude mapping with configurable range
  • 16 track groups with per-track solo/mute and color assignment
  • Cursor tooltip on hover shows time (or bar position in beat-sync) plus a color-coded dB reading per track; in Vertical mode the tooltip and cursor line restrict to the hovered lane
  • Peak hold with 2-second hold and 20 dB/s decay
  • Hold mode for phase alignment (shows last complete bar, swaps at boundary)
  • SIMD-optimized ring buffer with f32x16 mipmap reduction
  • Waveform renderer bypasses tiny-skia's raster pipeline — direct pixel-write column fills with half-split envelope smoothing (~52% less GUI CPU than the original path-based rasterizer)
  • CPU-rendered GUI with amber phosphor terminal theme (tiny-skia + softbuffer), freely resizable
  • 16 instances @ 2% CPU headless (0.13% per instance)
  • Per-track solo/mute/color controls with DAW track name via CLAP track-info
  • Inspired by PsyScope and RusovDmitriy/oscilloscope

Satch

A detail-preserving spectral saturator. Uses FFT-based spectral analysis to preserve quiet frequency components through the clipping process, producing textured flat-top clipping instead of featureless flat tops.

  • Independent Gain (input boost) and Threshold (clip ceiling) controls
  • Detail knob preserves quiet harmonics through clipped regions via per-bin spectral magnitude saturation
  • Knee crossfades between hard clip (0%) and soft tanh saturation (100%)
  • Clip-aware detail blend — only affects clipped portions, unclipped material is unchanged
  • CPU-rendered GUI (tiny-skia + softbuffer), freely resizable
  • ~0.82 MB RSS and 0.14% CPU per instance (100 instances @ 13.7% CPU, 82 MB total)
  • Inspired by Newfangled Audio Saturate

Six Pack

A six-band parallel multiband saturator inspired by Wavesfactory Spectre. Each band's EQ boost is computed against dry as a difference, then run through a per-band saturation algorithm — so a band at 0 dB contributes silence and gain effectively becomes the per-band drive.

  • 1 low-shelf + 4 peaks + 1 high-shelf, each with frequency / gain / Q controls
  • Six saturation algorithms per band: Tube, Tape, Diode, Digital, Class B, Wavefold
  • Per-band Stereo / Mid / Side routing applied to the diff before saturation
  • Linear-phase polyphase oversampling: Off / 4× / 8× / 16×
  • Global de-emphasis subtracts the linear EQ boost so only saturation harmonics remain audible
  • CPU-rendered GUI with EQ curve display + live spectrum analyzer overlay (tiny-skia + softbuffer), freely resizable
  • Inspired by Wavesfactory Spectre

Tinylimit

A low-latency wideband peak limiter for track-level use. Feed-forward topology with lookahead and dual-stage transient/dynamics handling.

  • Individual attack, release, knee, and transient controls
  • 7 built-in character presets (Transparent, Aggressive, Punchy, Smooth, Safe, Vocal, Loud)
  • Optional ISP (true peak targeting via ITU-R BS.1770-4)
  • Gain Link mode for auditioning limiting without loudness change
  • CPU-rendered GUI with input/output meters and GR readout (tiny-skia + softbuffer), freely resizable
  • ~1.0 MB RSS and 0.12% CPU per instance (50 instances @ 6.2% CPU, 50 MB total)
  • Inspired by DMG Audio TrackLimit

Warp Zone

A psychedelic spectral shifter/stretcher that transforms audio in the frequency domain using a phase vocoder. Makes familiar sounds alien -- voices from another dimension, instruments with impossible harmonic structures.

  • Shift (-24 to +24 semitones) for pitch shifting without time stretching
  • Stretch (0.5x to 2.0x) warps harmonic spacing for inharmonic/metallic textures
  • Freeze captures the current spectrum as a sustained drone
  • Feedback compounds spectral shifts for rising/falling Shepard tone effects
  • Low/High frequency range limits for selective processing
  • Scrolling spectral waterfall display with psychedelic color palette
  • CPU-rendered GUI (tiny-skia + softbuffer), freely resizable
  • 4096-point FFT phase vocoder with linear interpolation and phase-coherent bin remapping

Wavetable Filter

A wavetable-based audio filter that uses wavetable frames as FIR filter kernels. Load any .wav or .wt wavetable file and use its spectral content to shape your audio.

  • Raw Mode: Direct time-domain convolution (zero latency)
  • Phaseless Mode: STFT magnitude-only filtering (no pre-ringing)
  • 3D/2D wavetable visualization (click to toggle), real-time filter response with live input spectrum shadow
  • CPU-rendered GUI (tiny-skia + softbuffer), freely resizable; rotary dials show DAW modulation arcs and support right-click text entry
  • SIMD-optimized convolution (f32x16) with silence fast-path — idle plugins use near-zero CPU
  • Inspired by Kilohearts FilterTable and EB-FreakyTable

Build Requirements

  • Rust nightly toolchain (automatically configured via rust-toolchain.toml)
  • Linux system dependencies (see below)
# Install Rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# Ubuntu / Debian
sudo apt install libxcb1-dev libx11-xcb-dev libx11-dev libxcursor-dev \
  libasound2-dev libgl-dev libjack-jackd2-dev libwayland-dev

Building

# Build all plugins (VST3 + CLAP)
cargo nih-plug bundle gain-brain --release
cargo nih-plug bundle gs-meter --release
cargo nih-plug bundle hd26 --release
cargo nih-plug bundle imagine --release
cargo nih-plug bundle miff --release
cargo nih-plug bundle multosis --release
cargo nih-plug bundle nap --release
cargo nih-plug bundle pope-scope --release
cargo nih-plug bundle satch --release
cargo nih-plug bundle six-pack --release
cargo nih-plug bundle tinylimit --release
cargo nih-plug bundle warp-zone --release
cargo nih-plug bundle wavetable-filter --release

# Standalone binaries
cargo build --bin gain-brain --release
cargo build --bin gs-meter --release
cargo build --bin hd26 --release
cargo build --bin imagine --release
cargo build --bin miff --release
cargo build --bin multosis --release
cargo build --bin nap --release
cargo build --bin pope-scope --release
cargo build --bin satch --release
cargo build --bin six-pack --release
cargo build --bin tinylimit --release
cargo build --bin warp-zone --release
cargo build --bin wavetable-filter --release

Host-optimized builds

For local builds that exploit your CPU's SIMD (tiny-skia's AVX2 raster paths, auto-vectorized DSP code), wrap any cargo invocation with cargo xtask native. It detects AVX2+FMA+BMI2 on the build host and, when present, sets RUSTFLAGS=-C target-cpu=haswell for the child cargo process. On non-x86_64 or older x86_64 hosts it falls back to the default target-cpu.

cargo xtask native nih-plug bundle wavetable-filter --release
cargo xtask native build --release --bin gs-meter

The resulting binaries are NOT portable to pre-Haswell machines -- use plain cargo nih-plug bundle ... for distributable bundles.

The bundler outputs to target/bundled/. Copy either the .vst3 or .clap file (you only need one -- use whichever your DAW supports) to your plugin directory:

  • Linux: ~/.vst3/ or ~/.clap/
  • macOS: ~/Library/Audio/Plug-Ins/VST3/ or ~/Library/Audio/Plug-Ins/CLAP/
  • Windows: C:\Program Files\Common Files\VST3\ or C:\Program Files\Common Files\CLAP\

Workspace Structure

tract-plugin-pack/
├── gain-brain/             # Gain utility with group linking
├── gs-meter/               # Loudness meter + gain utility
├── hd26/                   # Unison-detune chorus + pseudo-stereo widener (Serum-inspired)
├── imagine/                # Multiband stereo imager modeled on Ozone Imager
├── miff/                   # MSEG hand-drawn FIR convolution filter
├── multosis/               # 16-row grid sequencer with per-row effects + MSEGs
├── nap/                    # EDVN velvet-noise character reverb with drawn curves
├── pope-scope/             # Multichannel real-time oscilloscope
├── satch/                  # Spectral saturator with detail preservation
├── six-pack/               # Six-band parallel multiband saturator
├── tinylimit/              # Wideband peak limiter
├── warp-zone/              # Spectral shifter/stretcher
├── wavetable-filter/       # Wavetable-based filter plugin
├── nih-plug-widgets/       # Legacy vizia widgets (no longer used; kept on disk for reference, excluded from the workspace build)
├── tiny-skia-widgets/      # Shared CPU-rendered widgets, editor base scaffolding
├── tract-dsp/              # Shared GUI-free DSP primitives (true-peak, FIR, STFT, ...)
├── docs/                   # Plugin manuals
│   ├── gain-brain/
│   ├── gs-meter/
│   ├── hd26/
│   ├── imagine/
│   ├── miff/
│   ├── multosis/
│   ├── nap/
│   ├── pope-scope/
│   ├── satch/
│   ├── six-pack/
│   ├── tinylimit/
│   ├── warp-zone/
│   └── wavetable-filter/
└── xtask/                  # Build tooling

Testing

cargo nextest run --workspace   # All tests -- parallel runner
cargo clippy --workspace        # Lint check

Install nextest via cargo install cargo-nextest --locked. Config lives in .config/nextest.toml.

Documentation

What is Clip-to-Zero?

License

GPL-3.0-or-later

Author

Michael Dungan