Audio Repair and Restoration

Recorded with noise suppression on? What can and cannot be fixed in REAPER

By StudioRescue Editorial

Published October 19, 2026

This is a different problem from the one most noise articles answer. The usual job is that there is too much noise in the file and you want less of it. Here, something already decided that for you, aggressively, in real time, and what you are left with is the result. Reaching for a noise reduction plugin at this point is the one move almost guaranteed to make it worse.

What it sounds like

Suppression damage has a recognisable set of symptoms. You will usually have several of these at once.

• The start of a word goes missing after a pause. You say "click on the start menu" and the file says "ick on the start menu". The detector needs a moment to decide that speech has begun, and the first consonant is spent on that decision.
• Quiet tails vanish. The ends of sentences fade out early, trailing words drop below the threshold and get cut off, and breaths are simply absent.
• The gaps are dead. Real silence in a real room is a thin band of low level noise. If your gaps are a perfectly flat line at zero, nothing in your room produced that.
• The background pumps. The room noise rises when you speak and disappears when you stop, so the noise floor breathes with your voice.
• A swirly, underwater quality on quiet sounds. Spectral processing that guesses wrong leaves a warbling artefact, most audible on sibilants, soft consonants and reverb tails.
• Anything that is not a voice suffers worst. A guitar, a jingle, a second person off mic. RØDE's own support article notes that on non-vocal material Windows 11's audio enhancements can cause a muffled sound or, in some cases, no sound at all.

If your recording simply has steady hiss or hum sitting in it and nothing is being chopped, you have the ordinary problem, not this one. Start with working out whether the hiss is your mic or your interface instead.

Where the processing came from

There are three layers between your microphone and your REAPER track, and any of them can process the signal without telling you.

1. The operating system

On Windows, digital signal processing is inserted into the audio path by components Microsoft calls audio processing objects. Microsoft's own developer documentation describes them as providing customisable software based digital signal processing for Windows audio streams, notes that the technology is symmetric and works essentially in reverse for input devices, and states that an endpoint effect is always applied, even to raw streams. In the Windows interface this shows up as an audio enhancements control on the properties of each input device. RØDE describes the feature as removing and dampening background noise while focusing on what the system detects as your voice, and makes the important point that the setting is applied per device, so switching it off for one microphone does not switch it off for another.

On macOS the equivalent is Mic Mode. Apple's instructions are to open an app that supports Mic Modes, click the control in the menu bar, then choose a mode: "To filter out background sounds, select Voice Isolation." Focusrite's support team is blunt about what this does to a recording through an interface. They write that when enabled it "can alter the audio quality of your input, and your channel mapping", that "'Voice Isolation' only works for speech" so instruments and synths may sound glitchy, and that "unless specifically required, we would recommend disabling this feature."

2. The microphone's own software

Most USB microphones and many interfaces ship with a companion app that includes a noise gate, a noise suppressor or a one click "broadcast" preset. These run before the signal reaches your DAW, and the processing is printed into the recording.

3. The app you recorded in

If the take came out of a conferencing or streaming app rather than REAPER, assume suppression was on unless you specifically turned it off. It is the default in most of them, because for a meeting it is the right default.

A two minute test that tells you for certain

1. Record about ten seconds where you say nothing at all, then count slowly to five, leaving a clear pause before each number.
2. Zoom in on the waveform in REAPER and look at the silent section. Real room tone is a thin but visible band. A dead flat line means something is gating or suppressing upstream.
3. Look at the start of each number. If the attack of each word is cut square rather than rising naturally, that is a detector, not your microphone.
4. Turn off every suppression switch you can find, record the same test again, and compare the two files side by side on separate tracks.

Do this test before an important session rather than after it. It takes less time than reading this paragraph.

Why it cannot be undone

Noise suppression works by deciding, moment by moment and frequency band by frequency band, how much of the signal is noise, then reducing the level of those parts. What gets written to your file is only the result. The decisions themselves, the amount of reduction applied at each instant in each band, are not stored anywhere. Without them there is nothing to invert. A gate is the same story in simpler form: when it closed, it wrote silence, and silence contains no information about what it replaced.

So be realistic about the ceiling here. Everything below improves how the damage reads to a listener. None of it recovers content.

What genuinely helps in REAPER

One optional move: suppression often shaves high frequency content it mistakes for noise, leaving the voice slightly dull. A gentle high shelf, somewhere in the region of +1 to +3 dB above roughly 6 kHz, can bring some presence back. Treat that as a starting point and not a setting, because the same shelf also lifts the warble, and on a badly processed file that trade is not worth making. Bypass and compare before you keep it.

What to avoid

• Do not run more noise reduction on it. The artefacts are not noise, and a denoiser handed a file with no stable noise floor to learn from tends to deepen the warble rather than remove it. If you want the honest picture on what those tools can and cannot do, see the free noise reduction plugins roundup.
• Do not add a gate to tidy up. The file has already been gated. Gating it again sharpens exactly the edges you are trying to soften.
• Do not reach for heavy limiting. It raises the artefacts along with everything else.
• Do not expect a restoration plugin to be the answer. Paid dialogue repair tools are better at smoothing artefacts than free ones, but they are still smoothing, not restoring, and the ceiling in the previous section applies to them too.

The pre-flight check, so it does not happen again

This is the part worth keeping. Before anything you care about, walk these four in order.

1. Windows: open the sound settings, go to the properties of the input device you are actually recording from, and turn off audio enhancements. Repeat for every input device you use, since the setting is per device.
2. macOS: open the microphone control in the menu bar while a recording app is running and set Mic Mode to Standard rather than Voice Isolation.
3. Your microphone's companion app: turn off the gate, the noise suppressor and any broadcast or voice preset. Record flat.
4. The recording app: in REAPER, there is nothing on by default, which is the point. In anything else, go and look.

Then run the two minute test above and confirm the gaps have room tone in them. If you are setting up a repeatable session, fold this into your routine alongside the rest of the podcast recording setup.

What this article will not fix

• It will not bring back missing syllables. Those were never written to the file.
• It will not restore a natural noise floor. A bed of substitute room tone is a convincing mask, not the original.
• It will not rescue non-voice material. Music or an instrument that went through voice focused suppression is usually beyond useful repair.
• It will not make a one take interview sound untouched. If words are lost and cannot be re-recorded, the honest options are a pickup line from the speaker, or a transcript correction, not a plugin.
• It is not the same as recording on the wrong device. If the take went to a laptop's built in microphone instead of your interface, that is a different rescue job, covered in what can be saved when you recorded on the laptop mic.