Find every quiet gap in an audio file
A silence detector scans the raw audio waveform and reports the start and end of every stretch that falls below a chosen loudness threshold. That is useful for trimming dead air from podcasts, splitting long recordings into tracks, or finding the boundaries between spoken phrases — all without uploading your file to a server.
How it works
The file is read into an ArrayBuffer and decoded to raw PCM with AudioContext.decodeAudioData, giving one Float32Array of samples per channel where each value sits between -1 and 1. For every sample the tool takes the absolute peak across channels and compares it to an amplitude threshold derived from the dBFS slider using amp = 10^(dB / 20). So -50 dBFS becomes an amplitude of about 0.00316.
Consecutive samples under the threshold are grouped into a run. A run only counts as silence if it is at least as long as the minimum-length setting, computed as minSamples = (minMs / 1000) * sampleRate. Each surviving run is converted to seconds with index / sampleRate to produce the reported timestamps.
Tips and notes
- Start at -50 dBFS and adjust: raise it toward -30 dBFS for noisy recordings, lower it toward -65 dBFS for clean studio audio.
- A minimum length of 300–500 ms avoids flagging the natural micro-pauses between words.
- The percentage figure (total silence ÷ duration) is a quick way to gauge how much dead air a recording contains before you edit.
- Because decoding happens in-browser, very large files use more memory; for multi-hour recordings consider analysing shorter sections.