core.io¶
Audio I/O and file utilities.
load_audio¶
Load an audio file.
Signature
load_audio(path, sr=None, mono=True) -> (audio, sample_rate)
save_audio¶
Save an audio file.
Signature
save_audio(path, audio, sr, source=None, actions=None, extra=None)
Also writes a provenance sidecar JSON at <path><suffix>.json.
write_sidecar¶
Write provenance sidecar metadata for any Triton-generated artifact.
Signature
write_sidecar(path, source=None, actions=None, extra=None) -> Path
sidecar_path¶
Return sidecar path for a generated artifact.
Signature
sidecar_path(path) -> Path
normalize_peak¶
Normalize audio to target peak amplitude.
Signature
normalize_peak(audio, target=0.99) -> audio
Scales audio so the maximum absolute sample value equals the target. Preserves dynamics.
Parameters:
audio(np.ndarray): Audio signaltarget(float): Target peak amplitude, 0.0–1.0. Default: 0.99
Returns: Normalized audio array
Use when: You need a safety headroom before mixing, or consistent peak levels across files.
normalize_rms¶
Normalize audio to target RMS (energy/loudness) level.
Signature
normalize_rms(audio, target=0.1) -> audio
Scales audio so the root mean square (RMS) amplitude equals the target. Duration-independent and perceptually meaningful.
Parameters:
audio(np.ndarray): Audio signaltarget(float): Target RMS amplitude, 0.0–1.0. Default: 0.1
Returns: Normalized audio array
Use when: You need consistent loudness across files of different lengths, or preparing audio for noise mixing (control SNR reliably).
Key difference from peak normalization:
RMS normalization is duration-independent. A 1-second file and a 10-second file with the same loudness pattern will normalize to the same amplitude regardless of length. This makes it ideal for datasets with variable-length files and for controlling signal-to-noise ratio consistently in mixing workflows.
rms¶
Compute root mean square.
Signature
rms(signal, axis=-1) -> float
is_audio_file¶
Check if path is supported audio file.
iter_audio_files¶
Iterate over audio files in path.
SUPPORTED_EXTS¶
Set of supported audio extensions.