Silence Finder

How the Silence Finder works

How the silence finder works

Silence is information. In a podcast it marks where the questions end; in a lecture recording it separates the slides; in a vinyl rip it divides the tracks; in a voice-over session it shows where the takes are. Finding those gaps by scrubbing through an hour of audio is tedious, and the tools that remove silence automatically do not tell you where it was. This one only looks — it changes nothing and reports everything.

How it measures. The file is mixed to mono and divided into 10 ms windows. For each window the tool computes the RMS level, the average energy of the window, which is a far better guide to "quiet" than the peak because a single click in an otherwise silent gap does not break it. A window is silent when its RMS is below the threshold; a run of silent windows that lasts at least the minimum length becomes a gap, reported with a start and end time to the nearest hundredth of a second.

Choosing a threshold. Digital silence — a gap inserted by an editor — sits at -infinity and any threshold finds it. Real room tone from a decent recording sits around -60 to -50 dBFS; a noisy laptop microphone might be at -40. The default of -50 dBFS works for most spoken recordings. If no gaps are found, raise the threshold towards -40; if the tool reports hundreds of tiny gaps between words, lower it or increase the minimum length. The Audio Info Inspector's RMS reading for the whole file is a useful reference point: silence should be well below it.

Minimum length filters out the natural pauses inside speech. A second is about right for finding the breaks between segments of a podcast; drop it to 0.3 s to catch every pause between sentences, or raise it to five seconds to find only the gaps between tracks of a recorded album side.

Leading and trailing. Two numbers most people want straight away: how much dead air there is before the recording starts and after it ends. They are reported separately so you know what to trim, and the Silence Trimmer will remove exactly that much.

Reading the list. Each gap is shown as a start and end clock time and a duration. The total silence line puts it in perspective — 8 % of an interview is normal, 30 % means long pauses that the Silence Remover will tighten. The page shows the first ten gaps; the export contains all of them.

Export formats. CSV gives an index, start, end and duration in seconds for each gap, suitable for a spreadsheet or a script that cuts the file at those points. Audacity labels produces a tab-separated text file in the format Audacity's Import Labels expects, so every gap appears as a named region on a label track under the waveform, ready to be turned into split points or checked by ear.

Related tools