Flues

Live synthesis experiments exploring physical modeling, speech synthesis, distortion algorithms, and hybrid techniques

---

Contents

• What is this?
• Try It Now - Live Web Experiments
• The Experimental Journey
  • Physical Modeling - Stove
  • Speech Synthesis - Chatterbox + ChatGen
  • Distortion Synthesis - Disyn
  • Hybrid Synthesis - Floozy + Flues-Synth
  • Drum Synthesis - Drumkit
• Implementation Patterns

What is this?

Flues is a collection of experimental synthesizers exploring different approaches to sound generation. What started as a simple clarinet simulator has evolved into a family of instruments spanning physical modeling, vocal synthesis, distortion techniques, and hybrid architectures.

Each concept begins as a browser-based experiment (using Web Audio API), then evolves through multiple implementations: native desktop apps (GTK4/C), DAW plugins (LV2), and embedded hardware (Raspberry Pi). The goal is to understand synthesis techniques deeply by implementing them across different platforms.

---

🎹 Try It Now - Live Web Experiments

No installation required - these run entirely in your browser:

Experiment	Synthesis Type	What It Does
Stove	Physical Modeling	12 interface types (reed, pluck, bow, brass, etc.) with waveguide synthesis
Chatterbox	Speech Synthesis	Formant-based vocal synthesis with IPA vowel quadrilateral control
Clarinet	Physical Modeling	Digital waveguide clarinet (the original experiment)

All web experiments support MIDI input, keyboard control, and touch/mouse interaction.

---

The Experimental Journey

1. Physical Modeling → Stove

Concept: Simulate acoustic instruments using digital waveguides, delay lines, and nonlinear interfaces.

Started with a clarinet synthesizer (Karplus-Strong algorithm), then generalized into Stove - a modular physical modeling engine with 12 different interface types representing physical (reed, bow, brass) and hypothetical (crystal, vapor, quantum, plasma) excitation methods.

Signal Flow: Sources → Envelope → Interface → Delay Lines → Filter → Modulation → Reverb

Available as:

• 🌐 Web app - Try in browser (docs)
• 🖥️ Desktop app - GTK4/C native (gtk-synth/)
• 🔌 LV2 plugin - For DAWs (lv2/pm-synth/)

2. Speech Synthesis → Chatterbox + ChatGen

Concept: Source-filter vocal tract modeling using formant cascade filters.

Chatterbox implements a larynx (voiced source), aspirator (noise), and four formant filters (F1-F4) representing jaw, tongue, lips, and voice quality. Includes vocal modes: nasal resonance, vibrato (sing), shout (boost), and vocal fry (subharmonics).

ChatGen is a companion LV2 plugin that converts typed English text into MIDI events controlling Chatterbox, enabling text-to-speech synthesis in a DAW.

Available as:

• 🌐 Chatterbox web app - Interactive vowel quadrilateral (docs)
• 🔌 Chatterbox LV2 - Full vocal synthesis plugin (lv2/chatterbox/)
• 🔌 ChatGen LV2 - Text-to-speech MIDI generator (lv2/chatgen/)
  • Usage: [ChatGen] → [Chatterbox] on same track

3. Distortion Synthesis → Disyn

Concept: Generate complex spectra using mathematical distortion functions.

Based on Victor Lazzarini's Distortion Synthesis tutorial, implementing 17 algorithms including: Dirichlet Pulse, DSF Single/Double, Tanh Square/Saw, PAF (Phase-Aligned Formant), Modified FM, and 10 extended algorithms with param3 control for enhanced spectral manipulation.

Available as:

• 🔌 LV2 plugin - Monophonic distortion synth (lv2/disyn/)

4. Hybrid Synthesis → Floozy + Flues-Synth

Concept: Combine multiple synthesis approaches in a single signal chain.

Floozy grafts Disyn's aggressive distortion algorithms onto Stove's physical modeling engine. Disyn provides the excitation source, feeding into the waveguide/filter/modulation chain for acoustic resonance.

Flues-Synth is the ultimate hybrid: a headless ALSA synthesizer for Raspberry Pi combining Disyn, formants, and physical modeling with MIDI program changes dynamically reconfiguring the signal chain.

Signal Flow: Disyn → Formants → Interface → Delays → Filter → Modulation

Available as:

• 🔌 Floozy (mono) - Single voice hybrid (lv2/floozy/)
• 🔌 Floozy Poly - 8-voice polyphonic (lv2/floozy-poly/)
• 🎛️ Flues-Synth - Headless Raspberry Pi (flues-synth/)
  • Signal flow diagram
  • Algorithm reference - Complete DSP documentation
  • 29 MIDI programs (0-28) switching signal chain configurations
  • 29 MIDI CCs + 6 control notes for live tweaking

5. Drum Synthesis → Drumkit

Concept: Synthesized drum voices inspired by TR-909 but pushed into hardcore industrial territory.

11 synthesized voices (kick, snare, clap, toms, hi-hats, crash, bash, cowbell, clave) using pitch envelopes, resonant filters, noise bursts, and inharmonic oscillators. The Euclid companion generates Euclidean/stochastic rhythms for Drumkit. Master FX chain includes bit crushing, distortion, and reverb for aggressive, metallic tones.

The Euclidean Gate experiment applies those rhythm concepts to any audio source by opening or muting audio with an AD envelope.

Available as:

• 🔌 LV2 plugin - 11 voices, 43 parameters, drum synth mapping (lv2/drumkit/)
• 🔌 LV2 plugin - Euclidean rhythm generator for Drumkit (lv2/euclid/)
• 🔌 LV2 plugin - Euclidean gate (audio effect) for rhythmic chopping (lv2/euclidean-gate/)

---

Implementation Patterns

Each synthesis concept follows a progression:

🌐 Web Experiment          → Learn the DSP, iterate quickly
   ↓
🖥️ Native Desktop App      → Optimize performance, direct hardware access
   ↓
🔌 LV2 Plugin             → Integrate with DAWs (Ardour, Reaper, etc.)
   ↓
🎛️ Embedded Hardware       → Deploy on Raspberry Pi / Daisy Seed

Technologies used:

• Web: JavaScript ES modules, Web Audio API, AudioWorklet, Vite, Vitest
• Desktop: C/GTK4, PulseAudio, Meson
• Plugins: C++, LV2, X11/Cairo UI, CMake
• Embedded: C/C++, ALSA MIDI/Audio, Meson

---

Getting Started

Try the Live Demos (Easiest)

Visit https://danja.github.io/flues/ and start playing immediately. All web experiments support MIDI controllers and computer keyboard input.

Download Pre-Built Releases

Pre-built binaries are available for Ubuntu desktop (x86_64) and Raspberry Pi (aarch64):

Download Latest Release

Three distribution options:

• Full Release (flues-v*.tar.gz) - All LV2 plugins + Flues-Synth
• Plugins Only (flues-plugins-v*.tar.gz) - LV2 plugins for desktop DAWs
• Synth Only (flues-synth-v*.tar.gz) - Headless synthesizer for Raspberry Pi

Quick install:

# Extract release
tar xzf flues-v0.1.0-multi-arch.tar.gz
cd flues-v0.1.0

# Install (requires sudo)
sudo ./install.sh

# Verify plugins
lv2ls | grep flues

# Run flues-synth (Raspberry Pi)
flues-synth

Build From Source

Prerequisites: cmake, build-essential, meson, ninja-build, lv2-dev, libx11-dev, libcairo2-dev, libasound2-dev

Quick start (builds all plugins):

git clone https://github.com/danja/flues.git
cd flues
./build_synths.sh --install-default

Plugins install to ~/.lv2/ and appear in any LV2 host (Ardour, Reaper, Carla, etc.)

Multi-platform release builds: See scripts/MULTI_PLATFORM_BUILD.md for building optimized binaries across desktop and Raspberry Pi platforms.

Individual builds: See subdirectory READMEs for specific instructions

• lv2/pm-synth/README.md - Stove plugin
• lv2/chatterbox/README.md - Speech synthesis
• lv2/chatgen/README.md - Text-to-speech
• lv2/disyn/README.md - Distortion synthesis
• lv2/floozy/README.md - Hybrid mono
• lv2/floozy-poly/README.md - Hybrid poly
• lv2/drumkit/README.md - Drum synthesizer
• lv2/euclid/README.md - Euclidean rhythm generator
• lv2/euclidean-gate/README.md - Euclidean gate effect
• lv2/quadrangle/README.md - Launchpad performance instrument
• lv2/padseq/README.md - Launchpad drum sequencer

Build Native Apps

GTK Desktop (Stove):

cd gtk-synth
meson setup builddir
ninja -C builddir
./builddir/pm-synth-gtk

Raspberry Pi (Flues-Synth):

cd flues-synth
meson setup builddir
meson compile -C builddir
./builddir/flues-synth    # Auto-detects MIDI/audio

---

Project Structure

flues/
├── experiments/           # Web-based experiments (JavaScript/Web Audio)
│   ├── pm-synth/         # Stove physical modeling
│   ├── chatterbox/       # Speech synthesis
│   └── clarinet-synth/   # Original clarinet experiment
├── lv2/                  # LV2 plugins for DAWs
│   ├── pm-synth/         # Stove plugin
│   ├── chatterbox/       # Speech synthesis plugin
│   ├── chatgen/          # Text-to-speech MIDI generator
│   ├── disyn/            # Distortion synthesis plugin
│   ├── floozy/           # Hybrid mono plugin
│   ├── floozy-poly/      # Hybrid polyphonic plugin
│   ├── drumkit/          # Drum synthesizer plugin
│   └── euclid/           # Euclidean rhythm generator
├── gtk-synth/            # GTK4 native desktop app (Stove)
├── flues-synth/          # Headless Raspberry Pi synthesizer
├── www/                  # Built static site (GitHub Pages)
└── docs/                 # Cross-project documentation

---

Philosophy

These are experiments, not polished products. The goal is to:

• Understand DSP algorithms by implementing them multiple ways
• Explore unconventional synthesis techniques (distortion, hybrid chains)
• Learn platform differences (web timing vs. native RT audio vs. embedded)
• Prioritize sound quality and expressiveness over preset count
• Keep code readable and educational (see CLAUDE.md)

Expect rough edges, experimental features, and ongoing evolution. That's part of the fun.

---

Recent Updates

v0.1.0 Release (December 2025)

First Official Release with multi-platform builds optimized for Ubuntu desktop (x86_64) and Raspberry Pi (aarch64).

Download: GitHub Releases

What's Included:

• 7 LV2 Plugins: Disyn, Chatterbox, ChatGen, PM-Synth, Floozy, Floozy-Poly, Drumkit
• Flues-Synth: Headless Raspberry Pi synthesizer with 29 MIDI programs
• Multi-Platform Binaries: Optimized builds for each target platform
• Complete Documentation: Algorithm reference, MIDI spec, build guides

Multi-Platform Build System:

• Desktop plugins built on Ubuntu with -march=native optimization (AVX, SSE)
• Raspberry Pi synth built natively on Pi 4 with ARM-specific optimization
• Three distribution formats: Full release, plugins-only, synth-only
• Build process documented in scripts/MULTI_PLATFORM_BUILD.md

Flues-Synth Expansion (29 MIDI Programs):

Expanded from 8 to 29 programs, leveraging 17 Disyn algorithms (7 original + 10 extended with param3) through creative signal chain routing:

• Programs 0-7: Original physical modeling presets
• Programs 8-17: Distortion synthesis explorations (ModFM Formant, DSF Bells, PAF Vowels, Hybrid chains)
• Programs 18-28: Experimental combinations (Morphing engines, feedback networks, spectral animators)

Each program includes optimized slider mappings (9 sliders) and level safety. See flues-synth/docs/PROGRAM_CHANGE.md for complete program guide.

Backwards Compatibility:

• OscillatorModule updated with optional param3 support (default 0.5)
• All existing code continues to work without modification
• New algorithms accessible via explicit param3 parameter

Algorithm Documentation:

• Comprehensive algorithms.md with implementation-level DSP detail
• Signal flow diagrams and level calibration notes
• Complete MIDI reference in flues-synth/docs/midi.md

---

Reference Materials

• CLAUDE.md - Project guidelines and development practices
• AGENTS.md - Agent collaboration notes
• Physical Audio Signal Processing - Julius O. Smith III
• Distortion Synthesis - Victor Lazzarini
• Wikipedia: Physical Modelling Synthesis, Digital waveguide synthesis, Karplus-Strong

---

Future Directions

• Hardware deployment: Port to Daisy Seed / Electro-Smith platform
• Polyphony: Add voice allocation to remaining plugins
• Preset system: Cross-platform preset storage
• WebAssembly: Compile C/C++ engines to WASM for web
• MPE support: Polyphonic expression for expressive controllers
• More hybrid experiments: What happens when formants feedback through delay lines?

---

License: See individual subdirectories Author: Danny Ayers Status: Active development, expect changes