Noise removal and noise gating are two different tools that podcasters often reach for interchangeably. They solve different problems, and using the wrong one makes the audio worse, not better. Here is how they work and when to use each.
What noise removal does
Noise removal -- sometimes called noise reduction or spectral noise reduction -- is a process that identifies consistent background noise in a recording and reduces or removes it. The typical workflow is to find a section of audio where only the background noise is present (no voice, no signal), sample that as a "noise profile," and then apply that profile across the entire recording to attenuate the noise frequencies while preserving the voice.
This works well for consistent background noise: HVAC hum, computer fan noise, street traffic at a stable level, appliance hum. These sounds have predictable frequency profiles that appear consistently throughout the recording. The noise removal algorithm can model them accurately and reduce them without significantly affecting the voice.
It works poorly for intermittent noise (a door slamming, a dog barking, someone coughing) and for noise that shares spectral characteristics with speech (voices from a TV in the next room, for example). For those, the algorithm has trouble distinguishing noise from signal and you either get incomplete removal or artifacts in the voice.
What a noise gate does
A noise gate is a different kind of tool. It does not analyze or model noise frequencies. Instead, it monitors the level of the audio signal and silences the output when the level falls below a threshold you set. When you are speaking, the signal is above the threshold and the gate is open. When you pause, the signal drops below the threshold and the gate closes, cutting to silence.
The noise gate does not remove noise from the audio. It replaces quiet sections -- the gaps between phrases, the silences between sentences -- with complete silence. Whatever noise was in those gaps disappears because the whole section is gated to silence. But the noise that appears during speech is completely untouched.
The result of gating without noise removal: the background HVAC hum disappears when you are not speaking, then snaps back when you start talking. This is often more distracting than continuous low-level background noise. The contrast between silence and hum draws attention to the hum in a way that constant low-level hum does not.
Gate artifacts: what goes wrong
Gating artifacts are the audible side effects of applying a noise gate improperly or too aggressively. They come in a few varieties.
Chattering: when a speaker's voice level fluctuates near the threshold -- during soft consonants, at the ends of sentences, during trailing phrases -- the gate may open and close rapidly, creating a stuttering or chattering effect. This is particularly common when the threshold is set too high.
Hard cuts: when the gate closes abruptly, the transition to silence is instant and jarring. Well-configured gates use attack and release times -- brief slopes at the open and close transitions -- to make the cuts less sharp. Gates with very short release times produce hard-cut artifacts; longer release times produce smoother transitions but may leave residual noise at the tail.
Noise pumping: the audible jump between gated silence and the noise that returns when the gate opens. This is the characteristic sound of heavy gating that does not also include noise reduction. You hear silence, then hum, silence, then hum -- aligned with the speaker's voice pattern.
The right tool for each problem
For constant background noise (HVAC, fan, steady traffic): use noise removal. This is what it is designed for. Apply it at a conservative setting first, listen to the result, increase only if needed. Over-applying creates artifacts in the voice more damaging than the original noise.
For clean silences between phrases: use a gate, but combine it with noise removal first. Remove the background noise, then gate the residual noise in the quiet sections. In this order, the gate closes on relative silence rather than on an already-active noise floor, and the artifact problem largely disappears.
For intermittent loud noise (singular events): neither tool handles this well automatically. Manual editing -- finding the noise event in the waveform or transcript, cutting around it -- is usually the better approach for loud intermittent events.
Gate settings for voice
For a typical voice recording in a home studio, a practical starting configuration: threshold around -55 to -50 dBFS (this closes on the noise floor while leaving comfortable headroom above normal speaking level), attack 5-10ms (fast enough to catch the starts of words cleanly), release 200-300ms (long enough for the gate to stay open through short pauses and word endings without cutting too early).
These settings will vary based on your noise floor and how loudly you record. The goal is a threshold that is comfortably below your softest voice and comfortably above your loudest noise. If those two ranges overlap -- if your softest voice is close to your loudest background noise -- neither setting will work cleanly, and fixing the recording environment is the more effective solution than adjusting gate settings.
The quick decision guide
Hear constant hum or hiss that runs through the whole recording: noise removal first.
Want clean silence in gaps between sentences: add a gate after noise removal, or let Reverbwell handle the silence trimming as part of its standard processing pass, which does not introduce the hard-cut artifacts of a manually-set gate.
Hear intermittent loud sounds that appear at specific moments: manual cut around the event.
Hear a voice that sounds like it was processed in a phone booth -- slightly hollow, sibilance artifacts, thin: noise removal was applied too aggressively. The remedy is to apply it at a lower strength setting; the aggressive setting removed noise but also took frequencies from the voice.
