Silence Detection

Detect and optionally trim silent regions from audio and video files.

Quick example

import { ffmpeg } from "ffmpeg-kit";

// Detect silent regions
const silences = await ffmpeg.detectSilence("audio.wav", {
  threshold: -40,
  duration: 0.5,
});

console.log(silences);
// [{ start: 0, end: 1.2 }, { start: 45.3, end: 46.1 }, ...]

detectSilence()

Find all silent regions in a file.

Function signature

function detectSilence(
  input: string,
  options?: DetectSilenceOptions,
): Promise<SilenceRegion[]>;

Options

Parameter	Type	Default	Description
threshold	number	-40	Silence threshold in dBFS
duration	number	0.5	Minimum silence duration in seconds

Return type

interface SilenceRegion {
  start: number;   // start time in seconds
  end: number;     // end time in seconds
  duration: number; // = end - start
}

trimSilence()

Remove silent regions from a file.

Function signature

function trimSilence(options: TrimSilenceOptions): Promise<TrimSilenceResult>;

Options

Parameter	Type	Default	Description
input	string	—	Input file path
output	string	—	Output file path
threshold	number	-40	Silence threshold in dBFS
duration	number	0.5	Min silence duration to remove
padding	number	0.1	Silence padding to keep around speech
trimStart	boolean	true	Trim silence at the start
trimEnd	boolean	true	Trim silence at the end
trimMiddle	boolean	false	Also remove internal silences

Result type

interface TrimSilenceResult {
  outputPath: string;
  originalDuration: number;
  outputDuration: number;
  removedDuration: number;   // originalDuration - outputDuration
  silenceRegions: SilenceRegion[];
  probeResult: ProbeResult;
}

Examples

Trim leading and trailing silence

const result = await ffmpeg.trimSilence({
  input: "recording.wav",
  output: "trimmed.wav",
  threshold: -50,
  duration: 0.3,
  trimStart: true,
  trimEnd: true,
  trimMiddle: false,
});

console.log(`Removed ${result.removedDuration.toFixed(1)}s of silence`);

Remove all silences (for condensed audio)

const result = await ffmpeg.trimSilence({
  input: "lecture.mp3",
  output: "condensed.mp3",
  threshold: -40,
  duration: 1.0,
  padding: 0.15,
  trimMiddle: true,
});

Detect silences first, then decide

const silences = await ffmpeg.detectSilence("interview.wav", {
  threshold: -45,
  duration: 2.0,
});

if (silences.length > 0) {
  console.log(`Found ${silences.length} long pauses`);
  // decide whether to trim based on content
}

Related

• Audio — normalize loudness
• Waveform — visualize audio amplitude