How to remove silences in Premiere Pro automatically (2026 guide)
The native way first (Text-Based Editing, Paper Edit), then plugins: dB threshold, minimum length and padding explained, preset values, prices checked 21 September 2026, and what silence removal cannot fix.
Premiere Pro removes silences natively through Text-Based Editing: transcribe the clip, open the filter in the Transcript panel, choose Pauses, then Delete All. It works without any plugin but offers no decibel threshold, no minimum length and no padding. Plugins cut on the waveform instead, which is what you want on music, noisy rooms and multicam.
I am a video editor and I built one of the plugins mentioned further down, so the native methods come first and their limits are stated plainly. If Premiere's own tools do the job on your footage, keep your money.
Does Premiere Pro remove silences automatically?
Yes, since Text-Based Editing shipped. Premiere transcribes the clip, shows the transcript in a panel, and cuts the timeline when you delete text from it. Pauses appear in the transcript as ellipses, and a filter lets you select and delete every one of them in a single action. Filler words — "um", "uh", "you know" — have their own filter.
The catch is in how it decides. Premiere works from the transcript, not from the audio level. It deletes what the speech model marks as a gap between words. You get no threshold in decibels, no minimum gap length, and no margin before and after the cut. On clean voice recorded in a quiet room, that is enough. On a room tone with a fridge in it, or on a clip with music under the voice, it is not.
Method 1 — delete pauses with Text-Based Editing (no plugin)
Five steps, around two minutes on a ten-minute clip.
- Open the Text panel.
Window > Text, then the Transcript tab. Select your clip on the timeline first. - Transcribe. Click Transcribe, pick the spoken language, and tick Separate speakers if there are several people in the recording. Premiere downloads a language pack the first time and runs the transcription on your machine.
- Open the filter. The funnel icon at the top of the Transcript panel. Choose Pauses. Every silence is highlighted in the transcript.
- Review before deleting. Not every pause is dead air. A breath before a punchline is pacing, not a mistake. Deselect the ones you want to keep.
- Delete. Delete removes one; Delete All removes every highlighted pause, ripple-shifting the timeline as it goes.
Repeat from step 3 with Filler words selected to strip the "ums" in the same pass. Adobe documents the procedure on its help page for detecting and deleting pauses; a good third-party walkthrough with screenshots was published by Juno School in April 2026.
What you should know before you run it on a real project: the deletions are real ripple edits. If you have already placed music, titles or B-roll on upper tracks, they will not follow. Do the silence pass first, decorate afterwards.
Method 2 — Paper Edit, for interviews (Premiere 26.5 and later)
Premiere 26.5, released in September 2026, added Paper Edit. Instead of deleting what you do not want, you select what you keep: you "select the sentences or dialogue you want to keep, and preview your selections before creating a new sequence" (Adobe, September 2026).
For a forty-minute interview that has to become six minutes, that is the faster direction of travel. Deleting silences from forty minutes leaves you with thirty-six minutes you still have to cut. Paper Edit gets you to a rough assembly in one pass. For a talking-head video that is already scripted, it is the wrong tool — there you want the silences gone and the rest untouched.
Method 3 — the manual way, and when it is still right
Widen the audio track, look at the waveform, and cut with C then ripple-delete with Shift+Delete. Nobody misses doing this on an hour of footage, but it stays the correct method in two cases: a thirty-second clip, where setting up anything else costs more time than the cuts; and any moment where the cut point matters musically, which no automatic tool will judge for you.
Method 4 — a plugin that cuts on the waveform
Silence-removal plugins work on the audio signal rather than the transcript, and they all expose the same three parameters. Understanding them is most of the job:
- Threshold, in decibels. How quiet a passage has to be to count as silence. Around −38 dBFS is a normal starting point for voice. A lower number (−45 dB) is more forgiving and keeps more; a higher one (−32 dB) cuts more aggressively and will start eating breaths.
- Minimum length. How long that quiet passage must last before it gets cut. This is the parameter that protects natural speech: set it to 0.2 s and you remove the gaps between words, which sounds robotic. 0.5 s and above only removes real pauses.
- Padding. A margin kept before and after each cut so the attack of a word is not clipped. Without it, everything sounds slightly swallowed.
These three settings explain almost every bad automatic cut you have ever heard. Too high a threshold plus too short a minimum length equals machine-gun speech.
What the presets actually contain
Presets exist because those three numbers are not obvious. In Black Hat Editor Tools, the four built-in presets set a threshold, a minimum length and a padding; the published threshold and minimum-length values are:
| Preset | Threshold | Minimum length | For |
|---|---|---|---|
| Podcast | −38 dB | 0.3 s | Two people talking, decent mics, natural rhythm |
| Interview | −42 dB | 0.5 s | Someone who thinks before answering; keeps the beats |
| Aggressive | −32 dB | 0.2 s | Short-form, vertical, maximum density |
| Gentle | −45 dB | 0.8 s | Noisy room, music underneath, or a calm narration you want to keep calm |
Each preset also applies its own padding, and you can save your own presets once you have found the numbers that suit your microphone. The AutoCut help page shows where they live in the panel.
Why the undo history matters more than you think
There are two ways to actually make the cuts, and they behave very differently a day later. A tool that drives Premiere's razor tool pushes one undo step per cut: an hour of footage can leave several hundred operations on the stack, and getting back to where you were means holding Ctrl+Z and watching the cuts come back one by one. A tool that rewrites the project file does the whole pass as one operation — our engine decompresses the .prproj, removes the silent segments, shifts everything and reimports, so a single Ctrl+Z reverses it all.
Which one you are holding is on nobody's feature list. Cut thirty seconds with it, press Ctrl+Z once, and look at the timeline. Do that before you trust any of these tools on a client project.
What it costs, across the market
Prices read from each vendor's own page on 21 September 2026.
| Tool | Cheapest plan that cuts silences | Free option |
|---|---|---|
| PremiereCopilot | Free (Smart Silences, daily limits) · Pro+ $6.39/mo yearly | Yes, no card |
| AutoCut | Basic $9.90/mo ($6.60/mo yearly) | 14-day trial, no card |
| FireCut | Starter $19/mo ($10/mo yearly) | 7-day trial |
| Phantom Editor | Free tier with a monthly AI allowance · Pro $19/mo yearly | Yes |
| AutoPod | $29/mo (Jump Cut Editor is part of the single plan) | 30-day trial |
| Black Hat Editor Tools | Essentiel $15/mo ($144/year) | Free plan, no card · 14-day Pro trial, card required |
If cutting silences is the only thing you want to automate, PremiereCopilot's free tier and AutoCut's $9.90 plan are the rational answers, and AutoCut's trial is the only one on this list that asks for no card at all. Our plan makes sense when the silence pass is one job among several — captions, B-roll, multicam, transcription — in the same panel. The full comparison of AI plugins for Premiere Pro lays that out per use case.
Which method for which footage
| Your footage | Use this | Why |
|---|---|---|
| Scripted talking head, quiet room | Text-Based Editing, Pauses + Filler words | Free, built in, good enough |
| 40-minute interview to cut to 6 | Paper Edit (26.5+) | You select what you keep, not what you drop |
| Podcast, two to eight mics | A plugin, Podcast preset (−38 dB / 0.3 s) | Per-track thresholds; the transcript route mangles overlapping speech |
| Voice over music or room noise | A plugin, Gentle preset (−45 dB / 0.8 s) | The transcript has no idea the music is not silence |
| A 30-second clip | By hand | Setting up anything costs more than the cuts |
| Vertical short, maximum density | A plugin, Aggressive preset (−32 dB / 0.2 s), then listen | Density wins on short form, up to the point it sounds like a machine |
What removing silences will not fix
A silence pass removes dead air. It does not remove the sentence you started three times, the tangent you regretted, or the "let me say that again". That is a different job — call it the derush — and it needs the transcript, not the waveform.
Most tools now ship something for it: AutoCut has AutoCut Repeat, FireCut folds repetition and filler removal into its Pro tier, and our panel does it from the transcript and AI analysis, which reads the whole transcript and marks the retakes and the tangents before it cuts anything. Whatever you use, review that pass. Silence detection is deterministic; retake detection is a judgement call, and it will sometimes remove the better take.
The short version
Try the native route first: Window > Text, Transcribe, funnel icon, Pauses, Delete All. It costs nothing and it is genuinely good on clean voice. Move to a plugin the day you need a decibel threshold, a minimum length and padding — that is to say the day you record in a real room, with music, or with several microphones. Then spend five minutes on the three parameters rather than trusting a preset blind, and check how the tool undoes before you run it on a client's project. If you are shopping around, the AutoPod alternatives comparison covers the multicam side of the same question.
Frequently asked questions
How do I remove silences in Premiere Pro without a plugin?
Select the clip, open Window > Text, click Transcribe, then use the funnel filter in the Transcript panel to select Pauses and click Delete All. Run it a second time with the Filler words filter to strip "ums". It is free and built into Premiere, but it works from the transcript, so it has no decibel threshold and no padding.
What dB threshold should I use to cut silences?
Start at −38 dBFS with a minimum length of 0.3 seconds for normal voice. Go to −45 dB / 0.8 s if there is room noise or music under the voice, and to −32 dB / 0.2 s only for short-form content where density matters more than natural rhythm. Always listen to thirty seconds before applying the pass to an hour.
Why does my footage sound robotic after cutting the silences?
The minimum length is too short. At 0.2 seconds you are removing the natural gaps between words, not the pauses between sentences. Raise it to 0.4–0.5 seconds and increase the padding so word attacks are not clipped. A threshold set too high (−30 dB and above) causes the same symptom by cutting into quiet syllables.
Can I undo automatic silence cuts?
It depends on the tool. Anything that drives Premiere's razor tool pushes one undo step per cut, so reversing an hour of footage means hundreds of Ctrl+Z. Tools that edit the project file directly reverse the whole pass in one step. Whatever you use, duplicate the sequence before the first run.
Does Premiere Pro's pause detection work in every language?
Pause detection follows the transcription, so it works wherever Premiere's Speech to Text supports your language — Adobe publishes the current list, and language packs are downloaded from within Premiere. Filler-word detection is more language-dependent than pause detection, since it relies on a list of known filler words.
Silence removal or transcript-based editing: which comes first?
Silences first, structure second, decoration last. Cut the dead air, then remove the retakes and the tangents from the transcript, then place music, captions and B-roll. Doing it in the other order means every ripple edit shifts assets you have already placed on upper tracks.