generate-audio-tests

Generates a battery of WAV test files covering all combinations of signal type, channel count, sample rate, and bit depth. Each file is accompanied by a JSON sidecar with signal metadata, and a manifest.json index is written at the end.

Synopsis

generate-audio-tests [OPTIONS]

Flags

Flag	Default	Description
`--out-dir`	`data_generated/test-audio`	Output directory
`--channels`	`1,2,6,16`	Channel counts (comma-separated; mono, stereo, 5.1, 9.1.6)
`--sample-rates`	`44100,48000,96000`	Sample rates in Hz (comma-separated)
`--bits`	`16,24`	Bit depths (16 or 24 only)
`--signals`	(all)	Signal types to generate (comma-separated; see below)
`--duration`	`10.0`	Duration in seconds (does not apply to `sweep`, which is fixed at 30 s)

Signal Types (`--signals`)

Value	Description
`id`	Per-channel identification tones — each channel gets a unique frequency (300, 600, … 6000 Hz). Useful to verify channel routing.
`thd1k`	Single 1 kHz sine at −3 dBFS. Standard signal for Total Harmonic Distortion measurement.
`thd100`	Single 100 Hz sine at −3 dBFS. Low-frequency THD measurement.
`imd_smpte`	SMPTE IMD test: 60 Hz + 7 kHz at 4:1 amplitude ratio.
`imd_ccif`	CCIF IMD test: 19 kHz + 20 kHz at equal amplitudes.
`sweep`	Logarithmic frequency sweep 20 Hz → 20 kHz, 30 s fixed duration. Use for impulse response and frequency response measurements.
`white_noise`	White noise (flat spectrum).
`pink_noise`	Pink noise (1/f spectrum, −3 dB/octave).
`m_noise`	M-weighted noise per ITU-R 468. Standard for acoustic measurement.

Output Layout

data_generated/test-audio/
  wav/
    id/
      id_ch1_sr44100_b16.wav
      id_ch1_sr44100_b16.wav.json   ← sidecar with signal metadata
      id_ch2_sr48000_b24.wav
      …
    sweep/
      sweep_ch2_sr96000_b24.wav
      …
    pink_noise/
      …
  manifest.json                     ← index of all generated files

Files that would violate Nyquist (e.g., a 20 kHz sweep at 22050 Hz sample rate) are skipped automatically and counted in the summary.

Examples

Generate everything (default):

generate-audio-tests

Stereo only, 48 kHz, 24-bit, sweep + pink noise:

generate-audio-tests \
  --channels 2 \
  --sample-rates 48000 \
  --bits 24 \
  --signals sweep,pink_noise

Custom output directory and duration:

generate-audio-tests --out-dir /tmp/audio-tests --duration 5.0

Sidecar Format

Each WAV file has a companion .wav.json with structured metadata:

{
  "format": "wav",
  "channels": 2,
  "sample_rate": 48000,
  "bits": 24,
  "duration": 10.0,
  "signal": {
    "type": "pink_noise",
    "description": "1/f spectrum, -3dB/octave (pink noise)"
  }
}

Use Cases

Engine regression tests — run the full plugin chain on known signals and compare output
Room measurement — use sweep files as the test signal when recording room impulse responses
Channel routing verification — use id tones to confirm each channel reaches the correct speaker
Codec/format testing — spot-check encoder/decoder fidelity across sample rates and bit depths