A good silence threshold sits a few dB above your recording’s noise floor and below your quietest speech. There is no universal number, because a home office, a lecture hall and a game stream all have different room tone. This guide explains each setting a silence remover uses, gives starting values by recording type, and shows how to check the result before you export.
The five settings every silence remover uses
Tools name them differently, but almost every silence remover works with the same controls.
| Setting | What it decides | Also called |
|---|---|---|
| Threshold | How quiet audio must be to count as silence | Level, sensitivity, silence level |
| Minimum silence duration | How long a quiet stretch must last before it is cut | Duration, minimum pause length |
| Padding | How much audio is kept before and after speech | Pre-roll and post-roll, handles, Truncate to |
| Minimum kept segment | The shortest piece of sound kept between two cuts | Minimum clip length |
| Click or blip handling | Whether short noises can interrupt a silence | Ignore clicks, transient rejection |
The threshold decides what counts as silence. Duration decides which silences matter. Padding, minimum kept segment and click handling decide how natural the cut sounds.
Threshold: start from your noise floor
The threshold is a level in dB relative to digital full scale. Anything below it counts as silence. The Audacity manual puts the rule simply: set it above the noise floor in the pauses and below the quietest speech. It also warns that a noisy recording may leave little or no room between the two.
To find your noise floor:
- Find a pause in the recording where nobody speaks. Room tone at the start of a take is ideal.
- Play it and read the level on your editor’s meter, or select it and look at the waveform scale.
- Note the typical level, not the occasional spike from a chair or a breath.
- Set the threshold about 4 to 8 dB above that level as a starting point.
If the room tone sits at -55 dB, a threshold around -48 to -50 dB is a reasonable first try. If the room tone sits at -40 dB because of a fan or air conditioning, the threshold has to go up to around -34 dB, and you need to check that soft speech still clears it.
Why you see -40 dB everywhere
-40 dB is a figure you will see recommended often. It works for many quiet home recordings with a decent microphone, but it is a guess about your room, not a standard. Audacity’s own current default is -20 dB, which treats a lot more audio as silence. On a quiet or unprocessed recording, soft syllables and word endings can fall under -20 dB and get cut. Measure first, then decide.
Threshold versus sensitivity
Some tools show a sensitivity slider instead of a dB value. It controls the same thing: how quiet audio must be before it counts as silence. Check the tool’s help for which direction cuts more. The fix for clipped words is the same either way: make detection less aggressive.
Minimum silence duration
This is how long the audio has to stay under the threshold before it counts. It is the setting that separates a dead pause from the natural gap between two words.
- Too short (under about 0.2 s for normal speech) and the tool cuts between words. Speech starts to sound rushed and breathless.
- Too long (over about 1.5 s) and only the longest gaps go, so most hesitation stays.
Audacity’s Duration defaults to 0.5 s and accepts values down to 0.001 s. Clipchamp’s silence removal, according to Microsoft, finds pauses longer than 3 seconds and leaves a small gap so the cut does not feel abrupt. Those two defaults show the range: Audacity tightens ordinary pauses, Clipchamp only removes clear dead air.
Padding: pre-roll and post-roll
Padding is the audio kept on either side of each cut. Without it, every cut lands exactly at the threshold crossing, which is often in the middle of a soft consonant or a decaying vowel.
- Pre-roll protects the start of a word. Sounds like “s”, “f” and “h” start quietly and can sit below the threshold for a few milliseconds before the rest of the word.
- Post-roll protects the end of a word. Vowels and final consonants fade out, and room reverb carries on after the voice stops.
A useful way to think about it: the gap left after a cut is roughly pre-roll plus post-roll. Audacity does not use padding. Its Truncate Detected Silence action shortens each silence to a fixed Truncate to length, 0.5 s by default, which plays the same role. Its other action, Compress Excess Silence, shortens each silence to a percentage of its length, 50% by default, so long pauses stay longer than short ones.
Minimum kept segment and click handling
Two smaller settings prevent the most common artifacts.
Minimum kept segment stops the tool from keeping tiny islands of sound between two silences. A cough, a mouse click or a chair creak in the middle of a long pause can split it into two silences with a 50 ms fragment in between. With a minimum kept segment of about 0.2 to 0.5 s, that fragment is removed along with the pause.
Click handling goes one step further. It lets short, sharp noises sit inside a silence without breaking it. Keyboard clicks in a screen recording and mouse clicks during a game are typical cases. Turn it off for music or sound effects with short hits you need to keep.
Starting values by recording type
These are starting points to tune by ear, not standards. Threshold is given relative to your measured room tone.
| Recording | Threshold | Min silence | Padding (each side) | Min kept segment |
|---|---|---|---|---|
| Podcast, quiet room, one mic per person | Room tone + 6 dB | 0.6–0.8 s | 0.15–0.25 s | 0.3 s |
| Talking-head video with jump cuts | Room tone + 6 dB | 0.3–0.4 s | 0.08–0.12 s | 0.2 s |
| Lecture or classroom | Room tone + 8–10 dB | 1.0–1.5 s | 0.25–0.3 s | 0.5 s |
| Gaming with music or game audio | Just below quietest speech | 0.8 s | 0.2 s | 0.4 s |
| Remote call recording | Room tone + 6 dB | 0.6 s | 0.2 s | 0.3 s |
A few notes:
- Podcasts need a natural rhythm, so keep padding generous. Two people talking over each other is not a silence problem.
- Talking-head video can take tight cuts, because the picture jump already signals the edit. The sister guide on jump cuts for talking-head video goes deeper.
- Lectures have long pauses for writing and thinking, and room noise is often high. See removing silence from a lecture recording.
- Gaming with music is the hard case. If music or game sound plays under your voice, the level rarely drops below a threshold set from room tone. Record your mic on its own track if you can, and detect silence on that track.
Multitrack recordings stay in sync
When each speaker has their own track, detect silence across all of them together, or the tracks drift apart. Audacity’s manual makes this point about its Truncate tracks independently option: it is off by default, and turning it on can pull separately recorded microphones out of sync. It also notes that the effect shortens the selection, so anything you had aligned to it needs checking afterward.
According to Adobe’s help, Adobe Audition handles this in Multitrack view through Effects > Diagnostics > Deleting Silence, which can detect levels from the audio and either strip the silences or split the clips so you can check each cut first.
How to check your settings
Do not judge a silence setting by the total time saved. Check it like this:
- Run it on a 2 to 3 minute sample that includes both fast talk and a long pause.
- Listen to the joins at normal speed. Clipped “s” and “t” sounds mean the threshold is too high or pre-roll too short.
- Look for rhythm. If speech sounds breathless, raise minimum silence duration or padding.
- Scan the waveform for leftover gaps that should have gone. Those point to a threshold below the noise floor.
- Then apply to the full recording, and spot-check the start, middle and end.
Common mistakes
- Copying a threshold from another recording. A setting that worked in one room fails in another.
- Removing every gap. Speech needs pauses. Cutting them all makes a voice sound rushed. See what dead air is for which pauses to keep.
- Normalizing after the fact without measuring again. Cutting silence changes integrated loudness slightly. The podcast loudness guide covers the order.
- Ignoring breaths. A loud breath can sit above the threshold and survive as a short segment. A minimum kept segment removes most of them.
Try settings on your own audio
The quickest way to learn these settings is to hear them on your own file. The free silence remover on this site detects silence in your browser, shows every gap, and lets you play or download the cleaned audio as WAV. It is free for recordings up to 3 minutes and the file never leaves your device. If you work in a desktop editor, the guides for Audacity and Premiere Pro show where the same controls live.
On iPhone and iPad, QuellCut starts from presets such as Podcast, Jump Cut, Gaming, Remote Call and Soft Voice, and exposes each setting from this guide: sensitivity, minimum silence duration, padding, minimum kept segment, ignore clicks and track noise. You review every cut on a timeline before export, and everything stays on the device.
Questions
- What is a good silence threshold in dB?
- There is no single right value. Measure the level of your room tone during a pause and set the threshold a few dB above it, while keeping it below your quietest speech.
- What is Audacity's default Truncate Silence threshold?
- The current Audacity manual lists a default Threshold of -20 dB, with a range of -80 to -20 dB, and a default Duration of 0.5 s.
- Why do my cuts clip the start of words?
- Either the threshold is too high, so soft consonants fall under it, or there is not enough padding before speech. Lower the threshold by a few dB or add pre-roll padding.
- Why is nothing being cut?
- Usually the threshold is below the noise floor, so the room tone never counts as silence. Background music has the same effect. Raise the threshold or work on a voice-only track.
- Should padding be the same before and after speech?
- It does not have to be. Word endings often fade out more slowly than they start, so many editors leave a little more room after speech than before it.
- How short can the minimum silence duration be?
- Audacity accepts values down to 0.001 s, but very short values cut the natural gaps between words. For speech, starting points between 0.3 and 1 s are more practical.
Sources: Audacity manual: Truncate Silence (current), Audacity manual: Truncate Silence (legacy), Microsoft Support: Silence removal in Clipchamp.