Podcast and Spoken-Word Audio
How to edit out filler words and breaths in REAPER
Published August 23, 2026
REAPER has no stock “filler word” detector, so every um, uh, and false start is a manual cut: split the item on either side of the word with S, delete the middle, and close the gap. Breaths are different. You usually reduce them rather than delete them, because speech with every breath removed sounds robotic. Turn Ripple editing on so gaps close automatically, use Dynamic split or Auto trim/split to strip long dead air, and lower loud breaths by hand or with a gentle ReaGate. All of this uses tools that ship with REAPER.
If you record yourself talking, your raw file is full of ums, restarts, gaps, and audible breaths. Cleaning them up is what makes a spoken take sound edited instead of live. The important thing to know before you start is that REAPER does not detect filler words for you the way some newer AI editors claim to. Filler words are removed by hand. What REAPER does automate is the boring part around them: closing the gaps, trimming long silences, and taming breath level. Here is a workflow that keeps the result sounding natural.
Three different problems, three different fixes
People lump this together as “clean up my talking,” but you are really dealing with three separate things, and each wants a different tool.
- Filler words and restarts (um, uh, you know, false starts). These carry no useful audio, so they get cut out completely. This is manual.
- Breaths. These are real and mostly natural. The goal is to lower the loud, distracting ones, not silence every one. Removing them all is a common beginner mistake.
- Dead air. Long gaps where nothing is happening. These can be trimmed in bulk with a stock action, which is where most of your time savings come from.
Set REAPER up first
Two settings make spoken-word editing far less painful before you cut anything.
Turn on Ripple editing
Click the Ripple editing button on the toolbar, or use the Options menu, and set it to Ripple all tracks or Ripple per track. With ripple on, when you delete a chunk of a take, everything after it slides left to close the gap automatically. Without it, every cut leaves a hole you have to drag closed by hand.
Enable automatic crossfades
In the Options menu, switch on “Auto-crossfade media items when editing.” Butt-joining two pieces of audio can click; a tiny crossfade at each edit hides it, and REAPER adds these for you as you cut. You can set the default crossfade length in Preferences under Project, Media Item Defaults.
Editing here is destructive to the arrangement, not the source file, but it is still smart to duplicate the track or save a versioned project first. That way an over-aggressive pass is one undo history away from safe.
Filler words: the manual cut
There is no shortcut around listening. Play the take, and when you hit an um or a restart, this is the loop:
Find the edges
Zoom in on the waveform. A spoken word is a clear burst; the gaps of near-silence on each side are your cut points. Working visually plus by ear is faster than either alone.
Split it out
Place the edit cursor just before the filler word and press S to split the item, then just after it and press S again. Now the um is its own item.
Delete and close
Select that middle item and delete it. With Ripple editing on, the following audio slides left and the gap closes. Alternatively, make a time selection over the filler and delete the selection, which also ripples closed.
Listen across the join
Play a second before and after the cut. You are checking that the rhythm of speech still sounds natural and that there is no click. Speech has a cadence, and a cut that is technically clean can still sound rushed if you removed the natural micro-pause along with the word.
Beginners often remove the tiny pauses between phrases along with the filler, and the result sounds breathless and unnatural. Leave a little space. If two words now bump together with no gap, the edit will draw attention to itself.
Dead air: strip it in bulk
Once the obvious filler is gone, you are usually left with long quiet gaps. You can trim these one at a time, but REAPER has two stock features that do it in a pass.
Dynamic split
Select the item, then open the Item menu and choose Dynamic split items. REAPER looks at the audio against a level threshold and marks everything below it as silence. Set the threshold so it sits above room tone and breath but below your speech, choose the option to remove the silent areas, and use the leading pad and trailing pad values to leave a little audio on each side of every word. Start with pads around 50–150 ms so word beginnings and endings are not clipped, and preview before you commit.
Auto trim/split
The action Item: Auto trim/split items does a similar job and is handy for trimming the silence off the head and tail of clips. Both tools are blunt instruments compared with cutting by hand, so they shine on obvious dead air and still want a listen-through afterward.
If a threshold-based split is eating the start of words or chopping soft consonants, your threshold is too high or your pads are too short. Raise the pads first, then lower the threshold. It is better to leave a hair of silence than to bite off a t or an s.
Breaths: reduce, do not delete
Breaths are part of natural speech. Silence every one and the voice sounds like a machine gasping in and out of existence. The realistic goal is to pull down the loud, wet, distracting breaths so they sit under the words instead of jumping out. Two stock ways to do it:
By hand (most control)
Select just the breath, either as its own split item or with a time selection, and lower it rather than cut it. Dragging an item’s volume handle down a few dB, or drawing a quick dip in a volume envelope, keeps a trace of the breath so the transition into the next word still feels human. This is slower but gives the most natural result, and it is what most careful spoken-word editors actually do.
With ReaGate (automatic)
REAPER’s stock gate, ReaGate, can duck quiet passages automatically. Add it to the track and set the threshold just above the breath level and below the speech level. Reasonable starting points are a threshold somewhere around -45 to -30 dB (set it by ear against your own recording), a fast attack of a few milliseconds so words are not clipped, and a release around 50–200 ms so it closes smoothly. Rather than let the gate slam fully shut, which sounds unnatural, blend a little of the dry signal back in so a hint of room and breath remains.
A gate reacts to level, not meaning. It cannot tell a soft-spoken word from a loud breath, so an aggressive gate will clip the front of quiet words and leave breaths that happen to be loud. Keep it gentle, use it as a helper on top of hand edits, and never as the whole solution.
A sensible order to work in
- Cut filler words and restarts by hand with Ripple editing on.
- Strip long dead air with Dynamic split or Auto trim/split, using generous pads.
- Tame loud breaths by hand, or with a gentle ReaGate, keeping some natural breath.
- Listen start to finish for rhythm, clicks, and anything that now sounds too tight.
Do the cutting before you worry about tone and loudness. Editing changes the timing and the level map of the whole take, so set your final volume afterward.
| Symptom | Likely cause | Test | Fix |
|---|---|---|---|
| “Um” and restarts everywhere | Normal raw speech | Play through and mark them | Manual split with S, delete, ripple closed |
| Long silent gaps between sentences | Thinking pauses, dead air | Look for flat waveform stretches | Dynamic split or Auto trim/split with 50–150 ms pads |
| Loud, wet breaths jumping out | Close mic, natural breathing | Solo and listen between phrases | Lower each breath by hand, or gentle ReaGate |
| Edits sound rushed or clicky | Cut too tight, no crossfade | Play across each join | Leave micro-pauses, enable crossfades |
What this will not do
Editing timing and breaths is not the same as cleaning the recording, so set expectations:
- It will not remove steady background noise like hiss, hum, or a fan. That is a separate job. See how to reduce background noise in a podcast in REAPER.
- It will not set your loudness. After editing you still need to level the voice and hit a target. See how to get a podcast to -16 LUFS in REAPER.
- It will not fix plosive pops on p and b sounds. Those need their own treatment. See how to remove plosives in REAPER.
- It cannot make a badly over-edited take sound natural again. If speech starts to feel clipped and breathless, you removed too much. Undo and leave more space.
Filler words are a manual cut in REAPER: split with S, delete, and let Ripple editing close the gap. Trim long dead air in bulk with Dynamic split or Auto trim/split using generous pads. Reduce loud breaths rather than deleting every one, by hand or with a gentle ReaGate, and always listen across your edits so the speech still breathes. Every step uses stock tools.
Sources
- Cockos, REAPER User Guide (current version 7.x, 2026) - item editing, splitting, Ripple editing, and Dynamic split of items. reaper.fm/userguide
- Cockos, ReaPlugs / ReaEffects - ReaGate stock noise gate and its threshold, attack, hold, and release controls. reaper.fm/reaplugs
- Sound on Sound, “REAPER: Speedy Podcast Editing” - stock REAPER workflow for cutting spoken-word takes. soundonsound.com
- Sound on Sound, “ReaGate Noise Reduction” - using ReaGate and blending dry signal to keep results natural. soundonsound.com