Cleaning dialogue

Remove silences: a complete guide

The Quiet Gap Remover finds the dead air in a recording and proposes cuts you review before anything happens. Done well, it tightens a talking-head edit without making you sound rushed. This guide covers the whole flow, from choosing a pacing to undoing the result.

How silence detection works

Um Out analyzes the audio on your chosen track and measures energy over time. Stretches that fall below a loudness threshold for long enough are marked as quiet gaps. This runs locally on the companion engine, so your footage does not leave your computer for this step.

Two ideas matter here: how quiet counts as silence, and how long a quiet stretch must last before it is worth cutting. The pacing presets set sensible values for both, and you can set them yourself later.

Step by step

Open your sequence and launch Um Out from Window → UXP Plugins → Um Out. Open Remove silences from the tool grid.
Under Choose the source, set Timeline area to Entire timeline, In / Out, or Selected clips, and set Voice track to the track carrying your dialogue — not music.
Under Choose the edit style, pick a Pacing: Relaxed keeps long beats, Balanced is a natural cleanup, and Fast removes every clearly detected pause. Fast is the default.
Press Find silences. Um Out shows a timeline map of every proposed cut and a transcript you can click through. Nothing has changed in Premiere yet.
Review the list. Play across it and untick any gap you want to keep.
Press the action button — it names the count, for example Remove 8 gaps. Um Out applies the approved cuts and verifies the result.

Fast is the default because it is what most talking-head footage wants, but it is the most aggressive of the three. On a recording where your pauses are part of the delivery, run Balanced first. It is easier to cut more later than to rebuild rhythm you removed.

Your original sequence is not touched

By default Um Out duplicates your sequence and applies the edit to the copy, leaving the sequence you had open exactly as it was. The copy appears in your project bin with the same name and a suffix. This is the Back up sequence before applying setting in Settings → Safety; turn it off if you would rather edit in place and rely on Undo.

Also in that section: Keep later clips at their timestamps. Off by default, which means the gaps close and your timeline gets shorter. Turn it on when other tracks are already synced to the timings and must not move — the quiet becomes empty gaps instead.

Keeping natural pauses

Not every silence should go. A beat before a punchline, a breath between ideas, or a deliberate pause for emphasis all read as “quiet” to a detector but matter to a viewer. Because Um Out shows every cut first, you keep these by unticking them in the review step.

Two settings help before you even scan, under Settings → Speech protection: Keep sentence-end pauses leaves a short beat after a complete thought, and Protect breaths stops a cut landing inside one. If dialogue and music share a track, Protect music on voice track is there too.

Cut style and audio smoothing

Both sit under Choose the edit style, next to Pacing.

Cut style decides what the audio does across each join. Your picture stays continuous on V1 either way.

Clean cut — a hard join at every removed gap. The default.
J-cut — the next clip’s audio starts early.
L-cut — the previous clip’s audio runs late.
Overlap — audio overlaps on both sides.

Audio smoothing is the crossfade at each join: No fade keeps the original volume, Short is 160 ms and prevents clicks and pops, Smooth is 350 ms and clearly audible. Short is the default and is the right answer for most dialogue. Reach for Smooth on roomy recordings or where a music bed runs underneath.

Exact timing controls

The pacing presets set the timings for you. To set them yourself, go to Settings → Timing and turn on Show exact timing controls. That reveals:

Shortest pause to cut — 300 ms by default. Lower it for more cuts; raise it to ignore short pauses entirely.
Room before speech and Room after speech, in milliseconds — how much quiet is held back on each side of the speech that is kept, so words do not start or end abruptly.
Speech sensitivity — higher hears more as speech, so fewer things are treated as silence.

These are part of the Local and Plus plans. On the free trial the section shows what it does and offers the upgrade rather than hiding it.

Reviewing and adjusting the result

After you apply, the cuts are ordinary edits on your timeline. You can trim any join by hand, restore a segment, or fine-tune a fade exactly as you would with edits you made yourself. Nothing is locked.

Undoing

Changed your mind? Press Cmd/Ctrl + Z immediately after the operation to step back through Premiere’s native edit history. Confirm the timeline after undoing before running a new analysis.

If a pass cut into words, switch Pacing to Relaxed and scan again, or raise Room before speech and Room after speech in Settings. You do not have to accept a proposal you do not like — the review screen has a Setup button that takes you back with your choices intact.

Tips for cleaner results

Clean audio detects better. If you use noise reduction, apply it before analyzing so room tone does not read as speech.
Work in sections on long recordings. Reviewing 10 minutes at a time is faster to judge than an hour at once.
Pair it with the Filler Cleaner afterward to catch the “um”s that are not silent.

Next, see how to connect your own API key for the AI-powered tools, or revisit how the free trial works.