dhvani

Core audio engine for Rust.

Buffers, DSP, resampling, mixing, analysis, synthesis, and capture — in a single crate. The audio equivalent of ranga (image processing) and tarang (media framework).

Name: Dhvani (ध्वनि, Sanskrit) — sound, resonance.

What it does

dhvani is the audio processing core — it owns the audio math so nobody else has to. Applications build their audio features on top of dhvani.

Quick start

[dependencies]
dhvani = "1"

use dhvani::buffer::{AudioBuffer, mix, resample_linear};
use dhvani::dsp::{self, Compressor, CompressorParams};
use dhvani::analysis;
use dhvani::clock::AudioClock;

// Create buffers
let vocals = AudioBuffer::from_interleaved(samples_a, 2, 44100)?;
let drums = AudioBuffer::from_interleaved(samples_b, 2, 44100)?;

// Mix
let mut mixed = mix(&[&vocals, &drums])?;

// Process
let mut comp = Compressor::new(CompressorParams::new()
    .with_threshold(-18.0).with_ratio(4.0).with_attack(10.0).with_release(100.0)
    .with_makeup_gain(3.0).with_knee(6.0),
44100)?;
comp.process(&mut mixed);
dsp::normalize(&mut mixed, 0.95);

// Analyze
let spectrum = analysis::spectrum_fft(&mixed, 4096)?;
let r128 = analysis::measure_r128(&mixed)?;
println!("Peak: {:.2}, LUFS: {:.1}", mixed.peak(), r128.integrated_lufs());

// Resample for output
let output = resample_linear(&mixed, 48000)?;

Features

# Everything (default: dsp + analysis + midi + graph + simd)
dhvani = "1"

# Core only — buffers, mixing, resampling, clock
dhvani = { version = "1", default-features = false }

# Media player — DSP + analysis, no MIDI or graph
dhvani = { version = "1", default-features = false, features = ["dsp", "analysis", "simd"] }

# Full synthesis + acoustics
dhvani = { version = "1", features = ["full"] }

Architecture

dhvani
├── buffer/        AudioBuffer, format conversion, mixing, resampling, dithering
├── clock/         Sample-accurate transport, tempo, beats, PTS
├── dsp/           Biquad, SVF, EQ, compressor, limiter, reverb, convolution, delay, automation, routing
├── analysis/      FFT, STFT, R128 loudness, dynamics, chromagram, onset/beat/key detection
├── midi/          MIDI 1.0/2.0, voice management, clips, routing, translation
├── graph/         RT-safe audio graph, topological execution, latency compensation
├── meter/         Lock-free peak/RMS/LUFS metering
├── capture/       PipeWire capture/output, device enumeration
├── simd/          SSE2/AVX2/NEON kernels with scalar fallback
├── synthesis/     Synth engines via naad (subtractive, FM, additive, wavetable, granular, drum, vocoder)
├── voice_synth/   Voice synthesis via svara (glottal, formant, phoneme, prosody)
├── acoustics/     Room acoustics via goonj (IR generation, convolution, FDN, ambisonics, presets)
├── creature/      Animal vocals via prani
├── environment/   Nature sounds via garjan
├── mechanical/    Mechanical sounds via ghurni
├── sampler/       Sample playback via nidhi
├── g2p/           Text-to-phoneme via shabda
└── ffi/           C-compatible API

Full details: docs/architecture/overview.md

Consumers

Dependency stack

dhvani (audio engine)
├── abaco (DSP math: amplitude/dB, poly_blep, panning, filters)
├── naad (synthesis engines)         [feature: synthesis]
├── svara (voice synthesis)          [feature: voice]
├── goonj (room acoustics)           [feature: acoustics]
├── prani (creature vocals)          [feature: creature]
├── garjan (environmental sounds)    [feature: environment]
├── ghurni (mechanical sounds)       [feature: mechanical]
├── nidhi (sample playback)          [feature: sampler]
├── shabda (grapheme-to-phoneme)     [feature: g2p]
└── bhava (personality/mood)         [feature: bhava-voice]

Building from source

git clone https://github.com/MacCracken/dhvani.git
cd dhvani

cargo build                          # default features
cargo build --features full          # everything
cargo build --features pipewire      # with PipeWire (Linux, requires libpipewire-dev)
cargo test --features full           # 663 tests + 22 doctests
cargo bench --features full          # 65 benchmarks

License

GPL-3.0-only. See LICENSE for details.