Remove the silence from your video

Your result, explained

The checker gives you three numbers: how many pauses it found, how many seconds of dead air they add up to, and how much shorter the clip is with them gone. A talking video that is a quarter shorter without its pauses is normal; a third is common for a first take. The waveform marks each pause so you can see where they cluster, usually at the start, before a hard line, and at the end.

Press play on the tightened version to hear the clip as it would sound cut. If a join sounds clipped, the pause was shorter than the breath on either side of it, and a longer minimum will leave it alone. If the clip still drags, lower the minimum and listen again.

What each setting means

Shortest pause to count: 0.3 seconds, 0.5 seconds or 1 seconds. Gaps shorter than this are breaths and are left alone. 0.5 seconds suits a fast short video; 1 second suits a calm one.

Background noise: quiet room, normal, or noisy. It moves the line between talking and silence. Pick noisy if there is a fan, traffic or music behind you, so the noise isn't counted as speech; pick quiet room for a clean recording, so soft word endings are kept.

How it works

The clip's sound is measured as one loudness reading every 20 milliseconds. From those readings the checker works out two levels: how loud your talking is and how quiet the room is when you stop. A pause is any stretch that stays near the room's quiet for at least the shortest length you chose. A click or a lip smack inside a pause doesn't end it.

When the pauses are skipped in the preview, 100 milliseconds are kept on each side of every cut, so the first and last letter of each line survive. The same room is kept when Krafto makes the cuts.

Because the clip is measured against itself, the settings work the same on a phone, a clip-on mic and a studio mic. The method only hears loudness: it can't tell a thoughtful pause from a dead one, it can't hear an "um" or a repeated line, and music under the voice hides the pauses altogether.

When this doesn't apply

Music or a sound bed under your voice: the pauses are never quiet, so none are found. Measure the voice track on its own.

Interviews and calm, emotional pieces: a pause is often the point. Use a longer minimum, or read the count as information and cut by ear.

Very noisy recordings: if the room is almost as loud as the talking, the checker has little to work with and may find nothing. Fix the sound first, or accept the count as rough.

The checker measures; it doesn't save a new file. To get the silence out of the video itself, make the cuts in your editor, or hand the clip to Krafto and ask.

Who made and checked this

The checker was built by the Krafto team, and its method and the numbers on this page were last checked on 7 October 2026. The made-up clip it shows before you add a file is measured by the same code as a real one, so the example can't promise anything the tool doesn't do. Spotted a clip it gets wrong? Tell us at hello@kraftoai.com.