Silero VAD Rust Implementation (vad-silero-rs)

A high-performance Rust implementation of the Silero Voice Activity Detection (VAD) model, optimized for real-time audio processing and integration into Rust-based applications. This project aims to provide an efficient and reliable solution for detecting speech segments in audio streams.

Table of Contents

• Silero VAD Rust Implementation (vad-silero-rs)
  • Table of Contents
  • Features
  • Installation
  • Usage
    • Basic Example
  • License
  • Acknowledgements

Features

• High Performance: Leverages Rust's performance capabilities for efficient VAD processing.
• Silero VAD Model Integration: Directly uses pre-trained Silero VAD models.
• Real-time Processing: Designed for low-latency voice activity detection in live audio streams.
• Easy to Integrate: Simple API for embedding VAD functionality into your Rust projects.
• Minimal External Dependencies (Runtime): Minimal runtime dependencies for streamlined deployment.
• Cross-Platform: Compatible with various operating systems supported by Rust.

Installation

Add vad-silero-rs to your Cargo.toml file:

[dependencies]
vad-silero-rs = "0.1.0" # Use the latest version

Or, if you prefer to clone the repository and build from source:

git clone https://github.com/EliFuzz/vad-silero-rs.git
cd vad-silero-rs
cargo build --release

Usage

The library provides a straightforward interface to initialize the VAD model and process audio chunks.

Basic Example

Here's a quick example demonstrating how to use vad-silero-rs to detect voice activity in an audio buffer:

use vad_silero_rs::{VadNode, SampleRate, VADProbabilities};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Initialize the VAD model with a specific sample rate.
    // Ensure your audio input matches this sample rate.
    let mut vad = VadNode::new(SampleRate::Rate16k)?;

    // Example audio chunk (replace with your actual audio data)
    // This should be a flattened array of f32 samples.
    // For a real application, you would load audio from a file or microphone.
    let audio_chunk: Vec<f32> = vec![0.0; 1024]; // 1024 samples of silence

    // Process the audio chunk
    let probabilities: VADProbabilities = vad.process_audio_chunk(&audio_chunk)?;

    println!("Voice probability: {}", probabilities.speech_probability);
    println!("Silence probability: {}", probabilities.silence_probability);

    if probabilities.is_speech() {
        println!("Speech detected!");
    } else {
        println!("Silence detected.");
    }

    Ok(())
}

License

This project is licensed under the Apache 2.0 License.

Acknowledgements

• The original Silero VAD project for providing the excellent VAD model.
• The Rust community for its robust ecosystem and tools.