@audio/loudness-sounds

Sound labeling — level-threshold region detection (Audacity Label Sounds)

npm install @audio/loudness-sounds

import sounds from '@audio/loudness-sounds'

Region detector modeled on Audacity's Label Sounds analyzer — not a loudness metric, but a level-threshold segmenter that ships in this family for the same input pipeline. 10 ms chunks are classed sound/silence against a level threshold, adjacent sound regions closer than minSilence are merged, regions shorter than minSound are folded into the next, and each surviving region is padded (pre/post) into its neighboring silence — never past a neighboring sound boundary, per the manual: "labels can overlap other labels, but cannot overlap previous or following sounds."

sounds(channels, { threshold: -30, minSilence: 1, minSound: 1 })

| Param | Default | | |---|---|---| | fs | 48000 | Sample rate, Hz | | threshold | -30 | dB level splitting sound from silence | | measurement | 'peak' | 'peak' (max |x|, most sensitive — catches low-duty-cycle transients), 'avg' (mean |x|), or 'rms' (sqrt mean x²) | | minSilence | 1 | Silence gaps shorter than this (seconds) are closed, merging the regions either side | | minSound | 1 | Regions shorter than this (seconds) are folded into the following region | | pre | 0 | Seconds to pad each region's start into leading silence | | post | 0 | Seconds to pad each region's end into trailing silence | | max | 10000 | Maximum number of regions returned |

Accepts Float32Array (mono) or Float32Array[] (multichannel — chunk level is computed across all channels together). Returns an array of { at, duration, label } (label = 'Sound N'), ordered by position.

Use when: auto-splitting a long recording into per-take/per-phrase regions, silence trimming, or generating label tracks for editors.

Modeled on Audacity's Label Sounds analyzer; a level-threshold segmenter, not an ITU-R/EBU loudness measurement.

Part of @audio/loudness — the loudness family umbrella. Documented from the reference implementation.

MIT © audiojs