Natural-language editing
Type or dictate an editing instruction and Clip Smasher carries it out on your own footage.
How to cut the silence out of a talking video
A ten-minute piece to camera is usually seven minutes of talking and three of thinking. Clip Smasher will find the quiet parts and cut them in one instruction. This is what it does, what it cannot know, and how to tidy up after it.
What it is doing
It is loudness detection, not speech recognition. The app reads the audio levels across the clip, finds the stretches that are quiet relative to the rest of it, and cuts them out.
It cannot tell a voice from a motorbike, and it does not know what you said. Understanding that is the difference between it working brilliantly and surprising you exactly once.
Three rules keep the result listenable:
- A gap has to last at least half a second to be worth cutting. The pauses between words and the breath before a sentence are shorter than that and survive, which is why the result still sounds like a person talking.
- A very short burst of sound is not kept. A click, a door, a single cough in the middle of a long silence is not a reason to keep a clip.
- Each kept section gets a little padding. Cutting exactly on the first sample of a word chops the front off it, so the cut sits slightly before the sound starts and slightly after it stops.
Doing it
Get the recording onto the timeline
Import the clip and drag it onto a video track. Levels are measured when the file is imported, so the analysis is already waiting by the time you ask for it.
Ask for it
Open the command panel and type or dictate the instruction. All of these work:
- chop out the parts where nobody is talking
- cut out the silence
- remove the dead air
- tighten up the pauses
- get rid of the quiet bits
- clean up the dead air
Watch it back
The cuts are applied straight away, all together as one step. Play the whole thing through before doing anything else.
If the result is obviously wrong for the recording, undo takes every cut back in one press. That is usually a sign about the audio rather than about the app, and the next section covers what to do.
Fix the joins by hand
Loudness cannot tell the difference between a thinking pause and the beat before a punchline. Step through the cuts with the frame chevrons and nudge any that landed badly. A short crossfade on the audio hides a join that clicks.
Cover the visible cuts
Removing silence from a locked-off shot leaves visible jumps where you moved between takes. The usual fixes: cut b-roll over the joins on a second video track, or lean into it and let the jump cuts be the style.
When it does not work, and why
| What you see | What is going on | What to do |
|---|---|---|
| It says there is nothing quiet enough | The recording is loud from end to end: usually constant background noise, air conditioning, traffic or wind. There is no contrast to find. | Apply noise reduction to the clip first, then run the silence removal again. |
| It says nothing is loud enough to keep | The whole recording is quiet. Often a microphone that was not connected, or a phone in a pocket. | Normalize or raise the clip volume first so there is a real difference between talking and not talking. |
| It kept the motorbike | Working as designed. It keeps loud things; it does not know which loud things are you. | Delete those pieces by hand afterwards. It is still far less work than cutting the silences yourself. |
| The speech sounds clipped | Rare, since short gaps are protected, but it happens on very fast delivery with hard consonants. | Undo, and instead trim the ends only with “top and tail every clip”, then cut the middle by hand. |
Just the ends
Often what you actually want is the dead air at the start and end gone and the middle left alone: the bit where you walked to the camera and the bit where you walked back to stop it. Ask for the edges specifically and the middle is untouched:
- top and tail every clip
- cut before the speaking starts
- trim after the talking ends
- cut it off the top
And once the piece is tight, the rest of the edit is the same sentence: instructions can be combined, so “chop out the parts where nobody is talking, add a crossfade between every clip, duck the music under the dialogue” is one command, one plan and one undo.
Silence removal questions
How does Clip Smasher find the silence?
By loudness. It reads the audio levels of the clip and finds the stretches that are quiet compared with the rest of it, then cuts those out and keeps what is left. It is level detection, not speech recognition: it can tell loud from quiet, but it cannot tell a voice from a passing motorbike.
Will it cut the gaps between my words?
No. A gap has to last at least half a second to count as silence worth cutting, because removing the natural pauses between words makes speech sound clipped and unnatural. Short breaths and beats between sentences survive.
What if my recording has background noise the whole time?
Then there is no contrast between loud and quiet and the app says so rather than cutting arbitrarily. Try noise reduction on the clip first, then run the silence removal again. If nothing in the recording is loud enough to keep, it tells you that too.
Can I just trim the dead air off the start and end?
Yes. Ask for the edges only and it leaves the middle alone: “top and tail every clip”, “cut before the speaking starts”, “trim off the tail”.
Can I undo it if it cuts too much?
Yes. The whole operation is applied as a single step, so one press of undo restores the clip exactly as it was.
Read next
Multi-track timeline
Video, audio, text and image tracks, trimming, splitting, stacking and frame-accurate navigation.
Make a VHS look on your phone
The twelve tape dials, what each one does to the picture, and three recipes to start from.
Tighten up your next piece to camera
Clip Smasher is free to download for iPhone and Android, and silence removal is in the free version.
Free to download · iPhone and Android · No account · Works offline