How to remove silence from a video automatically

Dead air is the easiest thing to fix in a talking video, and the easiest to fix badly. Every tool that does it is doing the same two things: deciding what counts as silence, then deciding how much of it to keep. Get those two numbers right and the result sounds like you just talk crisply. Get them wrong and it sounds chopped up and breathless.

Here’s how detection works, the settings I’d use and why, and the free and paid tools that do it, with the numbers each one actually uses (checked October 2026).

How automatic silence detection works

There are two approaches, and they fail in different ways.

1. Loudness threshold

The tool measures the audio level in short slices. Any stretch quieter than a threshold (in dB, or as a percentage of full volume) for longer than a minimum duration is marked as silence. Then it cuts that stretch, usually keeping a little padding (also called margin or buffer) either side so it doesn’t clip the sound next to it.

So you’re setting three numbers:

• Threshold. How quiet is “silent”. Too high (say −25 dB) and soft words get cut. Too low (−60 dB) and room noise counts as sound, so nothing gets cut.
• Minimum gap. How long a quiet stretch has to be before it’s cut. Too short and you start cutting the natural gaps inside sentences.
• Padding, or “shorten to”. How much of each pause survives.

auto-editor, Audacity, Filmora, Movavi and ffmpeg all work this way.

2. Gaps between words in a transcript

The tool transcribes the speech, times every word, and treats the space between one word ending and the next starting as the pause. Descript, Clipchamp’s auto cut, Tella and Blazing Video work this way. A humming fan doesn’t fool it, because the fan isn’t a word. The catch is that it’s only as good as the word timings: if the timings are loose, cuts land a little early or late.

Settings that sound natural, and why

These are for one person talking to a mic: tutorials, demos, voiceovers, screen recordings. Interviews and anything with music want gentler settings.

Setting	Start here	Why
Threshold	−40 dB (−35 dB in a noisy room)	Normal speech on a decent mic sits well above this, and a quiet room sits well below it. If quiet words go missing, lower it. If pauses survive, raise it.
Minimum gap	0.4 s	Gaps inside a sentence are mostly well under this. Go to 0.25 s and you start cutting into phrases. Go above 0.8 s and you only catch the long “where’s that tab” pauses.
Keep (shorten to)	0.15 s	Enough room for a word to finish and the next to start cleanly. Below about 0.1 s it starts to sound breathless. Above 0.3 s it barely feels edited.

A good way to set the threshold: find two seconds where you’re not talking, look at how loud that room noise is, and put the threshold about 10 dB above it. Every room is different, so a number from a blog post (including this one) is only a starting point.

Why shortening pauses beats deleting them

The obvious setting is “remove all silence”. It sounds terrible, for three reasons.

• Word endings are quiet. The end of “that’s”, “first” or “half” is a soft hiss that often falls below the threshold. Cut right at the threshold and words lose their last consonant.
• Breaths live in pauses. Take every breath out and the listener notices something is wrong, even if they can’t say what.
• Rhythm carries meaning. A slightly longer gap between points tells the viewer one idea has finished. Squash every gap to zero and your whole video becomes one long sentence.

Adobe’s own forum has a thread asking for exactly this: Premiere Pro’s pause deletion in text-based editing removes the whole pause, and users want a “reduce to” setting so a breath is left.

So the settings that matter are “gaps longer than X” and “shorten them to Y”. Most of the tools below have some version of Y, even if they call it margin or padding.

Free ways to do it

auto-editor (free, open source, Mac, Windows and Linux)

If you’re happy with Terminal, this is the best free option. On a Mac, install it with brew install auto-editor, or download the binary from GitHub. Then:

auto-editor recording.mp4

That writes a new file next to the original with the quiet parts gone. By default it uses --edit audio:threshold=0.04 (4% of full volume, about −28 dB) and --margin 0.2s, according to the project’s README.

The margin is added on both sides of each bit of speech, so a cut pause keeps up to 0.4 s. That’s a bit loose for a tutorial. I’d tighten it:

auto-editor recording.mp4 --margin 0.1s --edit audio:-40dB

Two features I’d miss in most paid tools:

• It can export a timeline instead of a video. Add --export resolve, --export premiere or --export final-cut-pro and you get the cuts as an editable project, so you can fix the few it got wrong.
• It can watch the screen as well as the mic. --edit "(or audio:0.03 motion:0.06)" keeps a section if there’s sound or movement, so it won’t cut the quiet five seconds where you’re showing something load.

Audacity Truncate Silence (free, audio only)

Effect › Special › Truncate Silence takes a threshold in dB, a minimum duration, and a “Truncate to” length, which is the shorten-don’t-delete approach built in. The manual also has a “Compress Excess Silence” mode that keeps a percentage of each pause, which sounds more natural on long pauses.

It only touches audio. Use it for a podcast or voiceover, never on audio you’ll put back under a video, because every cut moves the sound earlier and the lips stop matching.

Clipchamp auto cut (free, in the browser)

In Clipchamp, click the sparkle button on the timeline, then auto cut, then transcribe. It flags pauses and you can remove them all or one at a time. Microsoft’s help page says it only finds pauses longer than 3 seconds and leaves a small pause at each cut. There’s no setting to change the 3 s, so it catches the long stalls and nothing else.

ffmpeg (free, for scripting)

If you’re building your own pipeline, ffmpeg’s silencedetect filter lists the silent ranges and you cut around them yourself:

ffmpeg -i recording.mp4 -af silencedetect=noise=-40dB:d=0.4 -f null -

It prints a silence_start and silence_end for each gap. Turning those into an edit is up to you, which is why most people should use auto-editor instead.

Dedicated tools

Tool	Detects by	Control
Descript	Transcript word gaps	“More than” or “between” a length, shortened to a length you set
Tella	Transcript	Natural, Fast, Faster (800, 500, 300 ms)
Filmora	Loudness	Threshold 25%, minimum 0.5 s, 0.1 s buffer by default
Movavi	Loudness	Auto, or a manual threshold and minimum pause
Blazing Video	Aligned transcript word gaps	Longer than 0.4 s, shortened to 0.15 s by default, both adjustable

• Descript. Shorten word gaps in the AI tools panel lets you find gaps “more than” or “between” two lengths and shorten them to a target (their example is 200 ms), one at a time or all at once. It uses AI credits; the free plan includes 1 media hour and 100 AI credits a month at the time of writing.
• Tella. The Cut panel’s Silences options are Natural, Fast and Faster, which Tella’s own guide puts at pauses longer than 800, 500 and 300 ms.
• Filmora. Tools › Audio › Silence Detection, with a volume threshold of 25%, minimum length of 0.5 s and a 0.1 s buffer by default, per Wondershare’s guide.
• Movavi Video Editor. Silence removal with an Auto mode, or Manual where you set the threshold and minimum pause, and a “Without gaps” option (Movavi’s steps).

In Blazing Video (Mac)

Blazing Video is our Mac screen and webcam recorder, so this is the one I can describe from the inside. It transcribes on your Mac as you record, then aligns every word to the audio, so it knows exactly where each word ends. Cut Pauses in the AI tools panel uses those timings: every gap between words longer than 0.4 s is shortened to 0.15 s. You can find gaps from 0.2 to 3 s and shorten them to anything below that. It’s one undo step, and because it goes by words rather than loudness, a noisy fan doesn’t stop it working.

Run Cut fillers next to it to take out the ums and uhs, then press Update Link and the share link you already sent plays the tighter cut. More on that in editing a screen recording by deleting words. It’s free during the beta; download it here (macOS 15 or later).

Pitfalls

• Clipped word endings. The fix is padding. If words sound bitten off, raise the margin or “shorten to” by 0.05 s at a time, not the threshold.
• Background noise. A fan, air con or a noisy laptop can sit above your threshold, so nothing reads as silent. Raise the threshold, clean the audio first (noise reduction, then silence removal), or use a transcript-based tool.
• Music on the same track. If there’s a music bed under your voice, loudness tools never see silence. Remove silence before adding music.
• Quiet bits that matter on screen. In a screen recording, a silent five seconds might be the moment the thing you’re demoing actually happens. Check the cuts against the picture, or use auto-editor’s motion option.
• Audio-only cuts. Cutting the sound but not the picture drifts out of sync with every cut. Only use audio tools for audio.
• Fades. Audacity’s manual warns that Truncate Silence can eat the quiet tail of a fade. Remove silence first, add fades after.
• Not listening back. Play the result at 1× once, start to finish. Automatic cuts are right most of the time, and the one that isn’t is usually in the bit people watch.

FAQ

Can I remove silence from a video for free?

Yes. auto-editor is free and open source and runs on Mac, Windows and Linux. Clipchamp’s auto cut is free in the browser but only finds pauses over 3 seconds. Audacity is free too, but only for audio.

What dB level counts as silence?

There isn’t one right number because it depends on your room and mic. Start around −40 dB and set it about 10 dB above the background noise in a stretch where you’re not talking.

How long should pauses be in a video?

For tutorials and screen recordings, I’d shorten anything over 0.4 s to about 0.15 s. For interviews or storytelling, keep more, around 0.3 to 0.5 s, so it doesn’t feel rushed.

Does iMovie or QuickTime remove silence automatically?

Neither has an automatic silence removal feature, as far as I could find. You can split and delete by hand in iMovie, or run the file through auto-editor first and import the result.

Why do my cuts sound choppy after removing silence?

Usually because the padding is too short, so word endings and breaths are being cut. Keep 0.1 to 0.2 s of each pause rather than removing it entirely.