A forty minute recording of someone talking usually holds somewhere between six and twelve minutes of nothing: breaths, pauses, the gap after a sentence where the person is deciding what to say next. Cutting those out is the single most repetitive job in a talking head edit, and it is the one that decides how long the whole project takes.
The three ways, and what each one costs
By hand
- Total control over every cut
- Roughly one hour of work per ten minutes of footage
- Gets worse as you get tired, and the last twenty minutes are always the sloppiest
With a panel
- One pass over the whole clip
- Seconds instead of hours
- Only as good as the rule it uses to decide what counts as silence
1. By hand, with the razor tool
Put the playhead where the speech stops, press C for the razor, cut, move to where the speech starts again, cut, select the middle and ripple delete with Shift and Delete. Repeat a few hundred times.
This is worth doing at least once on a short clip, because it teaches you what a good cut sounds like: you want the breath before a sentence to survive, and you want the tail of the last word to survive, or the edit sounds clipped and anxious. Everything below is an attempt to automate that judgement.
2. With what Premiere already has
Premiere can transcribe your audio and, from that transcript, remove pauses in the Text panel. It works, it is free, and for a clean recording with one speaker it will get you most of the way. Two things limit it: the result depends on transcription quality, so an accent or background noise degrades the cuts, and the pause detection works on the transcript rather than the waveform, so a long breath that the transcript did not notice stays in.
Scene Edit Detection is a different tool and is often suggested by mistake: it finds cuts that already exist in a rendered file, not silences.
3. With a panel that reads the waveform
A silence removal panel looks at the audio itself and decides, sample by sample, where speech is and is not. Every one of them is built on the same basic idea and they differ almost entirely in one detail, which is the next section.
Why most silence cutters eat the ends of words
The obvious way to write one of these is to pick a volume threshold, call everything below it silence, and cut it out. The problem is what real speech looks like. A word does not end at full volume and then stop: it trails off. The letter s at the end of a word, the tail of a vowel, the consonant that closes a sentence, all of these sit below the threshold for a few dozen milliseconds.
The symptom, if you have seen it
The edit is technically correct and sounds wrong. Words end a fraction too early, sentences feel bitten off, and you find yourself dragging edit points back out by two or three frames over and over. That is a threshold cutting into the decay of the word.
FastEdit's Auto Cut is built the other way round. Instead of hunting for quiet and removing it, it finds the regions that are certainly speech and then grows them outwards until they meet silence that is unambiguous. A cut can only ever land in the space between two grown regions, so it cannot land inside a word.
Measured on real recorded speech against a fixed threshold method: zero cuts inside a word, against six for the threshold on the same clip.
Doing it in the panel
- Open the sequence you want to cut, with the audio on the timeline.
- Open the FastEdit panel from Window, Extensions.
- Pick the track to read. If your voice is on A1, that is the one.
- Set the padding you want kept before and after each phrase. A little padding is what makes an automatic cut sound like a human one.
- Run it. The gaps are removed and the remaining clips are closed up, so you do not get a timeline full of holes to tidy afterwards.
Everything it produces is ordinary Premiere cuts on ordinary clips. There is no proprietary object to maintain, nothing that breaks if the panel is uninstalled, and you can drag any edit point afterwards like any other.
This one is in the free tier
Auto Cut does not need a paid plan. Downloading the panel needs a free account, and after that there is no card, no countdown and no expiry on it.
Common questions
Does removing silence leave gaps on the timeline?
Not in FastEdit. The clips are closed up in the same pass, so the sequence stays continuous. Closing gaps is also available on its own, for timelines you cut by hand.
Will it cut inside a word?
It is built specifically so it cannot. The algorithm grows outwards from regions that are certainly speech rather than looking for quiet, so a cut can only land in the space between two phrases. Measured on real speech it made zero cuts inside a word, against six for a fixed threshold method on the same clip.
Does my audio get uploaded anywhere?
No. Silence removal runs entirely on your own machine. Only the AI features, which are a different part of the panel, send anything over the network, and the panel tells you before the first time it does.
Does it work on multiple speakers or several tracks?
It reads the track you point it at, so an interview with two microphones on two tracks is handled by choosing which track defines the speech. Cuts are applied across the sequence so the picture stays in sync.
Can I undo it if I do not like the result?
Yes. It is a normal Premiere operation and a single undo reverses it, so trying a different padding setting costs nothing.
Try it on the clip you are stuck on
Silence removal sits in the free tier and stays there. A free account to download, no card and no countdown.