A third-party noise filter does not plug into Microsoft Teams at all. It sits between your real microphone and Teams, pretending to be a microphone itself. You speak into your headset, the filter app cleans the signal, and it hands the result to Teams through a virtual audio device that shows up in the microphone list like any other piece of hardware. Understanding that one idea explains how to set these tools up, why they sometimes misbehave, and when they are worth the trouble.
The virtual device trick
Operating systems treat audio devices as interchangeable endpoints. An application asks for “the microphone” and receives a stream of sound, with no knowledge of whether it came from a USB headset, a built-in array, or a piece of software. Noise filter apps take advantage of this. When installed, they register a new input device with the system, and often a new output device as well.
The chain then looks like this. Your physical microphone feeds the filter app. The app runs its noise removal model on the audio in real time. The cleaned audio is published through the virtual microphone. Teams, set to listen to that virtual microphone, receives only the processed version.
The speaker side works in reverse. If you select the app’s virtual speaker in Teams, incoming audio passes through the filter before reaching your ears, which means it can remove the barking dog on a colleague’s end even though they have installed nothing.
Because all of this happens at the device level, the same virtual microphone can be selected in any calling or recording application, not only Teams.
Setting it up without doubling the processing
Setup always comes in two halves, and skipping either one is the most common mistake.
First, inside the filter app, choose your real microphone and real speakers as its sources. Second, inside Teams, open the device settings and choose the app’s virtual microphone, and its virtual speaker if you want inbound filtering. If you do only the first step, Teams keeps listening to the raw microphone and the filter does nothing. If you do only the second, the filter may be listening to the wrong physical device.
Then deal with Teams’ own noise suppression. Two filters in a row do not give you twice the cleanliness. The second filter receives audio that has already been shaped by the first, and the combination can produce clipped words or a watery, over-processed tone. The usual advice is to set the built-in suppression to Off or Low while the external filter is active, then make a test call from the device settings and listen to the playback.
Each product has its own quirks at this stage. For one widely used example, the setup notes on Krisp for Teams show the device selections on both sides and the matching Teams settings.
What the extra layer costs you
Real-time audio processing uses processor time, and it introduces a small delay because the app has to examine a short slice of sound before passing it along. On a modern computer neither is normally noticeable. On an older laptop already running video, screen sharing and a dozen browser tabs, the extra load can push things over the edge and cause stutter. Some tools move the work to the graphics card instead. NVIDIA Broadcast, for example, requires an RTX-class GPU, which rules it out for most business laptops but makes it attractive on a gaming or workstation PC.
A virtual device is also one more thing that can break. After an operating system update, or when a dock is plugged in, Teams may quietly switch back to a physical microphone, or the filter app may lose track of its source. If colleagues suddenly hear your background again, check the device selections in both places before anything else.
On a managed work computer, find out whether you are allowed to install the tool at all. These apps add audio drivers, which often needs administrator rights, and many require signing in to a vendor account. Security teams have legitimate questions about that, so ask before installing.
When built-in suppression is enough
For plenty of people, the suppression already in Teams handles everyday noise, and an additional layer adds complexity without much gain. An external filter becomes worth it in a few specific situations: you want identical filtering across several calling apps; you want to clean up incoming audio from other people; your organization has locked the built-in setting at a level that does not suit your environment; or you work somewhere truly loud, where comparing options side by side is sensible.
A quick recap before you install anything
A virtual microphone is a middleman. Point the filter at your real hardware, point Teams at the filter, turn down the built-in suppression so the two do not fight, and run a test call. If the recording sounds natural, you are finished. If it sounds processed, remove a layer instead of adding another.
