std.audio — PortAudio DSP & Mel Spectrogram Engine

Low-latency real-time audio capture, Cooley-Tukey Radix-2 FFT, Mel-scale filterbanks for Whisper speech AI integration, and Biquad IIR filtering.

1. Whisper AI Native Audio Pipeline (16 kHz Mono)

Configure low-latency PCM audio capture matched to the 16,000 Hz sample rate required by Whisper and speech models:

import std.audio.portaudio
import std.io

pub fn main() {
    // Initialize 16 kHz Mono audio engine
    let engine = portaudio.create_audio_engine(16000, 1, 512)
    std.io.println("Low-Latency Audio Engine Started: SampleRate=" + engine.sample_rate.to_string())

    // Convert acoustic frequencies to Mel scale for neural spectrograms
    let hz = 1000.0
    let mel = portaudio.hz_to_mel(hz)
    let roundtrip_hz = portaudio.mel_to_hz(mel)
    
    std.io.println("Frequency: " + hz.to_string() + " Hz -> " + mel.to_string() + " Mel")
    std.io.println("Roundtrip: " + roundtrip_hz.to_string() + " Hz")
}

2. Biquad Lowpass Audio Filtering

Remove high-frequency noise and alias artifacts prior to FFT computation:

import std.audio.portaudio
import std.io

pub fn main() {
    let mut filter = portaudio.create_lowpass_filter(16000.0, 4000.0, 0.707)
    let raw_sample = 0.82
    let filtered_sample = portaudio.filter_sample(&mut filter, raw_sample)
    
    std.io.println("Filtered Audio Sample: " + filtered_sample.to_string())
}