faust_min 0.5.4

High-fidelity physical modeling music synthesis engine powered by Faust, featuring 50+ dynamic instruments, polyphonic support, and an advanced UML multi-notation sequencer.

faust_min

A high-fidelity physical modeling music synthesis plugin and sequencing engine for Flutter, powered by Faust.

faust_min is a full-fledged, real-time computational acoustics engine based on Faust. Instead of relying on static, lifeless samples, faust_min mathematically simulates the physical properties of instruments (string tension, resonance, air pressure, tube length) to generate deeply expressive and highly realistic audio directly on the device. Result is very low memory and CPU footprint.

🌟 Key Features

🎸 50+ Physically Modeled Instruments

The engine ships with an extensive, highly optimized library of over 50 real-time physical models, covering a vast array of global instruments:

• Western: Electric/Acoustic Guitars, Pianos, Saxophones, Bass, Flutes, Drum Kits.
• Indian Classical & Folk: Sitar, Sarod, Santoor, Bansuri, Tanpura, Tumbi, Tabla, Dhol, Dholak.
• Cinematic Chinese Nature: Guzheng, Erhu, Dagu, Wind, Thunder, Chougong.
• Electronic & Ambient: Tibetan Singing Bowls, Rainmakers, Synths, Vocal Formant Synthesizers.

🎹 Advanced Polyphony & Multithreading

• Dynamic Voice Allocation: Built-in polyphonic support allows multiple notes to ring out simultaneously without cutting each other off. Availabe only for Piano and Guitars at the moment. Plan to add more western and eastern instruments in future.
• Persistent Thread Pool: Driven by a highly performant FaustMixer thread pool that renders heavy DSP algorithms concurrently across multiple CPU cores, avoiding UI thread blocks completely.
• Sample-Accurate Timing: Perfect for high-density, multi-instrument sequencing.

🎼 The Universal Music Language (UML) Sequencer

faust_min includes a deeply integrated, zero-latency Sequence Orchestrator capable of parsing plain-text text files into complex musical arrangements. The sequencer intelligently adapts to different cultural notations:

• Western Notation: Standard C4, D#5, Eb formatting.
• Indian Solfège (Sargam): Fully parses Sa, Re, Ga, Ma, Pa, Dha, Ni using complex Shruti frequency ratios.
• Percussion Bols/Strokes: Unpitched parsing of specific strokes like Dha, Ge, Dhin for Tabla or Kik, Snr, Hat for drum kits.

🚀 Getting Started

Installation

Add faust_min to your pubspec.yaml:

dependencies:
  faust_min: ^0.2.0

Initialization

Before creating instruments or sequences, initialize the engine and the audio mixer:

import 'package:faust_min/faust_min.dart';

void main() async {
  WidgetsFlutterBinding.ensureInitialized();

  // Extract the bundled DSP assets to app storage and configure the
  // native C++ engines. Unchanged files are skipped on subsequent launches.
  await FaustEngine.init();

  // Configure the mixer with a sample rate (48 kHz) and start the audio device.
  FaustMixer.instance.init(48000);
  FaustMixer.instance.start();

  // ... create instruments / sequences below
}

All instruments and sequences must be created with the same sample rate as the mixer (48 kHz).

Basic Instrument Usage

Generate high-fidelity audio on the fly:

import 'package:faust_min/faust_min.dart';

void main() async {
  WidgetsFlutterBinding.ensureInitialized();
  await FaustEngine.init();
  FaustMixer.instance.init(48000);

  // Create a new Flute instance (Instrument ID 10, StaticCompiled, 48 kHz)
  final flute = FaustInstrument.create(10, 0, 48000);

  // Play a note (Frequency, Velocity)
  flute.noteOn(freq: 440.0, velocity: 0.8);

  // Apply real-time physical parameter modulations
  flute.setParameter("breath_pressure", 0.9);
  flute.setParameter("vibrato_depth", 0.05);

  // Let it ring, then release
  flute.noteOff();
}

Advanced Multi-Instrument Sequencing

Orchestrate a massive arrangement using the UML Sequencer:

import 'package:faust_min/faust_min.dart';

void main() async {
  WidgetsFlutterBinding.ensureInitialized();
  await FaustEngine.init();
  FaustMixer.instance.init(48000);
  FaustMixer.instance.start();

  final orchestrator = SequenceOrchestrator();

  // Define a Tumbi sequence using Indian notation and parameter overrides
  String tumbiUML = '''
  grid: 4
  bpm: 120
  instrument: tumbi
  notation: Indian
  basefreq: 659.25
  vibrato_depth: 0.08
  vibrato_rate: 7.0

  5Sa . . 5Re . 5Ga . . 5Re . . 5Sa . . . 
  ''';

  // Define a Dhol sequence using Percussion strokes
  String dholUML = '''
  grid: 4
  bpm: 120
  instrument: dhol

  Dha . . Na Na . . Tin Na . . Dha Dha . . Na
  ''';

  // Load and play flawlessly synchronized
  orchestrator.addSequence("Tumbi", tumbiUML);
  orchestrator.addSequence("Dhol", dholUML);

  orchestrator.play();
}

🔨 Build Process (Native Engine)

The native engine must be built with CMake — there is no other supported build path.

Prerequisites

• CMake 3.10+
• A C++17 compiler (GCC/Clang)
• Faust compiler (faust) on PATH — the build auto-compiles every assets/dsp/*.dsp file into a generated Faust<Name>DSP.hpp header whenever the DSP changes
• FFTW3 (libfftw3) — required only by test_instruments

Configure

Run CMake from the linux/ directory (the CMake root that includes src/), into a build directory:

cmake -S linux -B build-release -DCMAKE_BUILD_TYPE=Release

Build — always in one shot

Whenever the faust_min core is modified (any .dsp, any src/*.cpp/*.hpp, or any assets/dsp/*.csv), all artifacts must be built together in a single invocation — never build only one artifact:

cmake --build build-release -j$(nproc)
# or equivalently
make -C build-release -j$(nproc)

This is mandatory because every test binary (test_sequence, test_mixer, test_instruments) links against the faust_min shared library, and faust_min post-build syncs all DSP CSV lookup tables. Building one test in isolation leaves the others linked against a stale core — the whole set (engine + all test binaries) must be rebuilt together so they stay in lockstep. CMake tracks the generated DSP headers automatically, so this is safe and incremental.

Artifacts

• build-release/shared/libfaust_min.so — the engine (bundled by the Flutter Linux plugin).
• build-release/shared/test_instruments, test_mixer, test_sequence — test binaries (also reachable via build-release/).
• DSP calibration CSVs are auto-synced to both assets/dsp/ and build-release/assets/dsp/ on every build.

Running tests

Test binaries resolve libfaust_min.so via the build dir, so run them from the repo root with LD_LIBRARY_PATH set:

LD_LIBRARY_PATH=build-release/shared ./build-release/test_instruments <id> f=220 v=0.8 s=0 --render

🏗 Architecture & Stability

Legacy C++ audio libraries often suffer from thread-safety issues, SIGSEGV crashes, and timing drift when used in a modern, multithreaded environment like Flutter's Dart Isolates.

faust_min solves this. The underlying DSP code is generated by Faust (Functional Audio Stream), which outputs completely stateless, re-entrant, memory-safe C++ arrays that are fully decoupled from the UI. This ensures industrial-grade stability and zero latency for professional audio applications.

🚀 Roadmap

• Enhanced Playing Techniques: Support for enhanced playing techniques for all instruments.
• Enhanced Mixer: Support for chaining (parallel/serial) various DSP elements like echo, bandpass, etc.
• More wrappers: Currently only Dart but in future Java, and Python too.

📄 License

This project is licensed under the MIT License.