EN
Download

Guides

Remove silence from video, automatically and without the robotic feel

Pauses, ums and false starts often take up a big chunk of a raw take. Cut well, they vanish without anyone noticing the edit. Here are the settings that work.

Updated 5 min readOrbicut team

Before · 0:42
After · 0:31
8 silences · 3 “ums” cut
Preview: Orbicut removes the silences and ums from a 40-second take

Why remove silences from a video?

On camera, nobody talks like a news anchor. We search for words, breathe, restart sentences. In conversation, those pauses go unnoticed; in a short video watched on a phone, they make people swipe away.

In an unscripted take, pauses and hesitations can easily fill a third of the running time or more: a 60-second take often becomes a 40-second video that is denser and easier to watch to the end. It is the first step in editing any talking video, before captions and music.

Step by step

  1. Import the take

    Import the raw video with its original audio, before adding music or effects.

  2. Set the threshold

    Pick a minimum silence length: 0.3 to 0.4 s for tight pacing, 0.5 to 0.7 s by default, 0.8 s or more for a calm tone.

  3. Set the padding

    Keep 0.1 to 0.2 s of silence before and after each phrase so syllables aren't clipped.

  4. Remove fillers and false starts

    Delete ums and uhs and keep only the best version of any repeated sentence.

  5. Hide the joins

    Add a slight zoom on every other cut or a cutaway over visible jump cuts.

  6. Watch it through

    Play the video from start to finish and restore any intentional pause that was removed.

Recommended settings by video type
Video typeMin. silence lengthPadding around speech
Quick TikTok or Reel tip0.3 to 0.4 s0.08 to 0.12 s
Standard talking head0.5 to 0.7 s0.1 to 0.15 s
Tutorial with demo0.6 to 0.8 s0.15 s
Customer testimonial, interview0.8 to 1.2 s0.15 to 0.2 s
Podcast clip0.7 to 1 s0.15 to 0.2 s
Sensitive or emotional topic1 s or more0.2 s

What exactly should you cut?

  • Silences: any pause longer than your threshold (see below).
  • Filler words: "um", "uh", "er", and in many cases "like" or "you know" when they add nothing.
  • False starts: "So the, well, the first thing...". Keep only the complete version of the sentence.
  • Repeats: when you said the same line twice, keep the best take, usually the last one.
  • Head and tail: the moment you hit record and the moment you reach for the phone.

Keep, however, deliberate pauses before a key point or punchline, and natural breaths between ideas. A video with zero breathing room quickly becomes exhausting.

What settings should you use?

Two settings drive most of the result.

Minimum silence length

This is how long a pause must be before it gets cut:

  • 0.3 to 0.4 s: very tight pacing, typical of TikTok and quick-tip Reels.
  • 0.5 to 0.7 s: dynamic but natural, a good default for most talking-head videos.
  • 0.8 to 1.2 s: relaxed pacing for testimonials, interviews, video podcasts or sensitive topics.

Padding around speech

This is the bit of silence kept before and after each phrase. With 0.1 to 0.2 s of padding, word starts and endings stay intact. Below 0.05 s you hear clipped consonants ("...ight" instead of "right") and the voice sounds choppy.

If your tool detects silence from volume, a third setting matters: the decibel threshold. Too low and background noise counts as speech; too high and whispered sentence endings disappear. Test on 20 seconds before processing the whole video.

Waveform detection or transcript detection?

There are two automatic methods:

  • Waveform-based: the tool cuts everything below a certain volume. Simple and fast, but a loud "um" counts as speech, and background noise (air conditioning, traffic) can block detection.
  • Transcript-based: the tool transcribes speech word by word with a timestamp for each word. It knows exactly where each word starts and ends, catches filler words and false starts, and cuts between words rather than mid-breath.

The transcript method gives cleaner cuts on spoken content, especially in slightly noisy rooms. The waveform method is still useful for videos without speech, such as a tutorial set to music.

How to remove silence manually

You can do it all by hand in any editor:

  1. Show the audio waveform and zoom into the timeline.
  2. Spot the flat areas between blocks of speech.
  3. Cut just after the end of one phrase and just before the start of the next, leaving about 0.1 s of padding.
  4. Delete the gap and close it up (ripple delete).
  5. Listen to every join: if you hear a clipped breath or a click, nudge the cut a few frames.

Expect 5 to 10 minutes per minute of video. It is a good way to learn, but it gets tedious fast if you post several videos a week.

Hiding the cuts: jump cuts and zooms

When you cut a talking-head take, the head "jumps" slightly at each edit: that is a jump cut. On social media it is accepted, even expected, but you can smooth it:

  • A slight zoom on every other cut (from 100% to about 110%): the framing change looks intentional and resets attention.
  • A cutaway (B-roll) over the join: the product, the job site, the screen.
  • Text or a caption appearing right on the cut pulls the eye elsewhere.

Our guide to auto zooms and jump cuts covers these techniques in detail.

Common mistakes

  • Cutting too tight: the video feels sped up and leaves viewers breathless. When in doubt, raise the minimum silence length.
  • Cutting the music with the voice: add music after removing silences, on a separate track, or it will jump at every edit.
  • Shifting background noise: if room noise changes from cut to cut, apply noise reduction first so the joins disappear.
  • Skipping the final watch: an automatic pass can drop a short word or an intentional pause. Watch the whole video once.

Removing silence with Orbicut

Orbicut automatically cuts silences, filler words, false starts and scene changes using a word-by-word transcript. You set the minimum silence length and the padding kept around speech; every cut shows on the timeline and can be undone, because nothing is destroyed. The app also adds a slight zoom on every other cut to hide the joins. Try it in the online demo or download Orbicut.

For the next step, see our animated captions guide.

Frequently asked questions

Does removing silence make a video feel fake?

Only when cuts are too tight or have no padding. With a minimum silence length around 0.5 seconds and 0.1 to 0.2 seconds of padding, most viewers won't notice the edit.

Can I remove silence from a video with several speakers?

Yes, but use a longer threshold, around 0.8 seconds, to keep the exchange natural. Check that short reactions like "yes" or "right" weren't removed.

Should I remove breaths?

Not all of them. Removing loud gasps tightens the video, but removing every breath makes the voice sound mechanical. Padding of 0.1 to 0.2 seconds keeps natural breathing.

Does silence removal work in other languages?

Yes. Waveform detection doesn't depend on language, and transcript-based detection works in every language the speech model supports, including each language's own filler words.

Are automatic cuts permanent?

It depends on the tool. In Orbicut, every cut stays visible on the timeline and can be reversed; the original video is never modified.