Remove Silence from Audio: Auto-Detect and Trim Pauses

Stop manually scrubbing through long recordings to find silent gaps. DuneTools analyses your audio, finds every silent stretch, and removes them with adjustable threshold and minimum length. Perfect for podcasts, lectures, voice memos and screen recordings.

100% Private

dB Threshold

Visual Waveform

No Watermark

Setup

How Auto-Silence-Removal Works

Three steps, all in your browser via Web Audio API.

Drop your audio

Drag-and-drop an MP3, WAV, OGG or M4A. The tool decodes it locally with the browser's Web Audio API, no upload involved.

Tune detection settings

Set silence threshold (default -30 dB, lower for noisy recordings, higher for clean ones), minimum silence length (default 0.5 s) and padding (keep 100 ms before/after for natural pauses). Preview the waveform live.

Export the cleaned audio

Download as WAV (lossless) or MP3 (compact). See exactly how much you saved (Original 18:42 -> Cleaned 14:21, 23% shorter).

Designed for Spoken Audio

Podcasts, lectures, voice memos, transcription prep, screen recording cleanup.

Adjustable Threshold

Default -30 dB works for most clean recordings. Push it lower (-40, -50 dB) for very quiet rooms. Push it higher (-20 dB) for noisy environments where 'silence' still has background hum.

Min Silence Length

Default 0.5 seconds. Set to 0.2 s to remove micro-pauses (aggressive), or 2 s to only kill long awkward gaps (gentle). Natural speech pauses are preserved.

Padding Safety

Keep 100 ms before and after each cut to avoid jarring jump-cuts. Adjustable 0-500 ms. Without padding, audio can sound abrupt as speech resumes.

Visual Waveform Preview

Canvas waveform with silent regions highlighted in red, kept regions in green. See exactly what will be cut before exporting.

100% Local Processing

Web Audio API decodes and analyses your file entirely in the browser. Audio data never leaves your device, ideal for confidential interviews, medical recordings or NDA content.

MP3 + WAV Export

WAV for lossless quality (drop straight into your DAW). MP3 via ffmpeg.wasm for compact files ready to publish or share.

Silence Detection: Technical Notes

Understanding the algorithm so you can tune it for your audio.

What 'silence' actually means in dB

Audio loudness is measured in decibels relative to full scale (dBFS), where 0 dB is the loudest the file can be without clipping, and -infinity dB is true digital silence. Real recordings have a noise floor: a quiet office is around -50 dB, a treated studio around -70 dB. Speech peaks at -6 to -12 dB. We use RMS over 50 ms windows to decide if a chunk is silence, RMS smooths out brief pops that would fool peak detection.

If your threshold is -30 dB but your room hums at -25 dB, nothing gets detected as silence. Lower threshold or denoise first.

Why min silence length matters

Natural speech has micro-pauses (50-300 ms between words) that you do NOT want to remove, those are how humans parse sentences. Setting a 0.5 s minimum means only stretches of 500 ms or longer get cut. This preserves cadence while removing the dead air between paragraphs, slides or false starts.

Podcast editors typically use 0.4-0.8 s min length. Aggressive sport edits use 0.2 s. Documentary work uses 1-2 s.

Padding: avoiding jump cuts

Without padding, you get jarring jumps: word ends -> instant new word -> sounds unnatural. Padding keeps a buffer of audio on each side of the cut (default 100 ms). The breath or trailing sibilant is preserved, the cut blends naturally. Set higher (200-300 ms) for music or careful narration, lower (50 ms) for fast-paced edits.

Listeners typically cannot tell silence was removed when padding is ≥80 ms. Below that, fades or crossfades start being necessary.

When this tool is NOT enough

Silence removal is great for content trimming, not noise removal. If your audio has continuous background hiss (fan, hum, traffic), this tool cannot remove it, you need a noise reducer or denoiser. Likewise, breath sounds, mouth clicks and 'um's are not silence, they are quiet noise that lives above the threshold. For that level of editing, a DAW (Audacity free, or Descript paid) handles it.

Use this tool first to remove dead air, then a DAW for noise reduction and mouth-noise cleanup if needed.

Real-world use cases

Podcast post-production

Cut every 'um, let me think' pause across a 90-minute episode in seconds. Saves 1-2 hours per episode versus manual scrub-and-cut.

Online lecture cleanup

Recorded a 2-hour class with 20 minutes of 'let me find that slide' dead air. Strip those gaps before sharing, students get a tighter playback experience.

Voice memo cleanup

iPhone voice memos with long thinking pauses, perfect for posting to a podcast feed or sending to a transcriber. Cleaner audio = faster transcription = lower cost.

Screen recording audio

Tutorials and demos often have 5-10 second pauses while tools load. Strip those before publishing to YouTube or your help center, viewers get straight to the content.

Interview prep

Field interview with awkward gaps while the source thinks. Auto-trim so transcripts and quote pulls flow naturally.

Audiobook self-recording

Self-narrating an audiobook? Strip the unintentional 'paragraph break' gaps before chapter export. Listeners get a professional flow.

Silence Remover, Frequently Asked Questions

Quick answers about the tool

What is the best threshold setting?

Start with the default -30 dB. If too much real audio gets cut, raise the threshold toward -20 dB. If silent pauses persist after processing, lower to -40 or -50 dB. Watch the waveform preview, red regions are what will be removed.

Will this remove breaths between sentences?

Probably not, breaths usually sit at -20 to -30 dB which is above default threshold. If you do want to remove them, lower threshold to -15 dB AND set min silence length to 0.3 s. Test on a 30-second sample first.

Can I use this for music files?

Technically yes, but it is designed for spoken audio. Music has continuous waveforms with very few true silent gaps (only between tracks). For music, the only useful pass is removing extra leading or trailing silence on a single track.

Is this lossless?

WAV export is lossless (the kept audio is bit-identical to original input). MP3 export is lossy (re-encoded at 128 kbps default). If quality matters, export WAV and convert to MP3 separately with your preferred bitrate using the Compress Audio tool.

How long can my audio file be?

Limited by browser RAM. A modern laptop comfortably handles 2-3 hours of speech audio (around 100-200 MB MP3 / 1-2 GB WAV). On mobile, stay under 1 hour for best performance.

Does it work on noisy recordings?

Partially. If background noise is above your threshold, nothing gets detected as silent. Try raising threshold toward -20 dB. For truly noisy material, run a denoiser first (Audacity has Noise Reduction free) then come back here to trim the silent gaps.

Can I see what will be removed before exporting?

Yes. The waveform preview shows the entire audio file with silent regions highlighted in red and kept regions in green. You see exactly what the algorithm decided, adjust threshold/min-length until the marking matches your intent, then export.

Will the timing of my speech change?

Yes, silent gaps are removed, so absolute timestamps shift earlier. If you have a transcript or subtitle file synced to the original, you will need to re-sync after silence removal. For interviews, this is usually fine. For lectures with slide changes timed to audio, plan accordingly.

Is my audio file uploaded?

No. Everything runs in your browser tab via Web Audio API (decoding) and ffmpeg.wasm (MP3 export). Zero outbound requests during processing, verifiable in DevTools.

What if no silence is detected?

The tool shows a clear notice and lets you download the original file unchanged. Usually means your threshold is too low (try -20 dB) or your audio truly has no silent gaps.