Audio foundations
What is noise suppression?
Noise suppression continuously separates your voice from steady background noise and strips the noise away, even while you are talking. Here is how it works and how it differs from a simple noise gate.
The short answer
Noise suppression is processing that listens to your microphone, figures out which part of the sound is your voice and which part is steady background noise, and removes the noise while keeping the voice. The key word is continuously. Good noise suppression cleans the signal the whole time, including the moments while you are actually speaking, so the fan hum, the air conditioner, the distant traffic, and the computer whir sit underneath your words and get taken out from under them.
That is what makes it different from just muting your mic between sentences. Suppression does not wait for a gap. It works on every moment of audio and pulls the noise out of the voice itself.
Suppression versus a noise gate
People often mix up noise suppression with a noise gate, and the difference matters. A noise gate is a simple on-off switch. When your voice is loud enough it opens and lets sound through; when you stop talking it closes and mutes everything. A gate is great at cleaning up the silence between words, which is why a quiet room sounds quiet with a gate on.
But a gate does nothing about noise while you are talking. When the gate is open, the fan hum rides right through alongside your voice, because the gate has no idea how to tell them apart. It only judges loudness, so it either passes the whole signal or mutes the whole signal.
- A noise gate mutes the gaps. Between words you get real silence, but during speech the background noise is still there.
- Noise suppression removes the noise under your voice too. It separates voice from noise moment to moment, so the fan is gone even mid-sentence.
Input
hiss everywhere
Suppression
gone under the words
Gate
stays under the words
The two are complementary, not competitors. A gate handles the pauses cheaply, and suppression handles the noise buried in the parts where you are actually speaking. Used together they give you a clean room and a clean voice.
How AI noise suppression works
Older noise reduction guessed at the noise by sampling a quiet moment and subtracting that profile from the rest. It worked for a constant hiss but struggled with anything that changed, and it often left your voice sounding thin or underwater when it overreached.
AI noise suppression takes a different approach. A model is trained on huge amounts of speech mixed with every kind of noise, so it learns what human voice looks like and what noise looks like. In real time it runs your audio through that trained model, which recognizes the voice, holds onto it, and rejects the rest. Because it has learned the shape of speech rather than a single static noise profile, it can pull out keyboard clatter, a barking dog two rooms over, or an uneven hum without hollowing out your voice.
This is a form of real-time digital signal processing, where the effect runs on every block of audio as it streams past. The model does more work than a plain filter, but a well built one keeps the added delay small enough that live calls, streams, and recordings feel immediate.
What noise suppression does and does not fix
Suppression is aimed at steady, ongoing background noise: fans, air conditioning, computer and GPU whine, room tone, distant road noise, the hum of a busy space. That is exactly the noise a gate cannot touch while you talk, and it is where suppression shines.
It is not a cure for everything, and honesty here saves you frustration. Suppression will not fix a room that echoes, because reverb is your own voice bouncing back rather than separate noise. It will not undo heavy clipping from a mic set too loud, and it is not a substitute for good mic placement. If you are chasing an echo instead of a hiss, that is a different problem, covered in how to reduce background noise on your mic. Suppression works best on a signal that was reasonable to begin with.
Noise suppression in patchd
In a virtual audio mixer the natural place for suppression is right at the input, before your voice goes anywhere else. patchd gives every channel a free effects rack, and AI noise suppression is one of the effects in it, included on the Free tier, on every channel. You add it from the + Add Node menu, put it early so a clean signal feeds everything downstream, and reorder it against the other effects by dragging.
It ships with an honest Mix control rather than a single on switch. Aggressive suppression can start to affect the natural tone of your voice, so the Mix knob lets you blend between your raw signal and the fully cleaned one and stop at the point where the noise is gone but you still sound like yourself. That tradeoff is real, and giving you the dial instead of hiding it is the point.
Because patchd is a real-time mixer, the cleaned voice then leaves on a bus that other apps pick up as a microphone: your call app, game, or stream software selects Patchd Bus 1 in its mic picker and hears the suppressed version, never the noisy input. Pairing suppression with the rack's noise gate covers both cases at once: the gate quiets the pauses, and suppression removes the noise living under your voice.
Getting the most out of it
A few habits make any noise suppression sound better. Start with the cleanest input you can: move the mic closer to your mouth so your voice is louder relative to the room, and the model has an easier job telling voice from noise. Set the Mix so the noise disappears without your voice turning brittle, and trust your ears over pushing the effect to maximum.
Order matters too. Suppression early in the chain means everything after it, including any compression, EQ, or a voice changer, is working on already clean audio instead of amplifying the hiss. Get the noise out first, then shape the sound. Do that, and the person on the other end hears your voice and nothing else, whether you are in a quiet booth or a room with a fan running.