Teams uses Krisp? Record both noise-control stages

You turn off noise suppression in Teams, but a sound you wanted to transmit still changes or disappears. Before choosing another microphone or enabling more features, inspect the input path. Teams may be receiving a Krisp virtual microphone whose physical source and processing have their own settings.
Start with the content you intend to send: ordinary speech, an instrument demonstration or another specific sound. Keep the result tied to that task. A configuration intended to remove background sound from speech may answer a different question from preserving material for a music lesson.
Name the source at each stage
Use the blank processing-stage record to keep the physical microphone, Krisp choice and Teams input beside each other. Record the app and operating-system versions, and leave any stage you cannot inspect unresolved.
| Stage | Record | Separate question |
|---|---|---|
| Physical source | Microphone selected inside Krisp | Is that the source you intended? |
| Krisp processing | Relevant microphone-side cancellation choice | What processing is enabled before Teams receives it? |
| Teams input | Named input shown in Teams | Is it Krisp or another source? |
| Teams processing | Noise and music choices actually available | What configuration suits this session's content? |
| Receiving endpoint | What an informed participant hears | Does the transmitted result match the task? |
This is an inventory, not proof that each stage is active or working correctly. A familiar microphone name alone cannot document the whole path.
Read Krisp's setup in its own scope
Krisp's Teams setup guide describes selecting Krisp Microphone or Speaker in Teams, then choosing physical devices within Krisp. It separately names microphone-side and other-participant noise cancellation.
The maker recommends turning Teams noise suppression off for its intended Krisp setup and explicitly states that doing so does not affect Krisp's audio processing. It also recommends keeping Teams high-fidelity music mode off for that setup, if available. Record which component owns the choice before treating one disabled switch as a bypass of everything upstream.
Use the help matching your installed versions if labels differ. This inspection does not require installing Krisp, buying a plan or replacing an existing microphone.

Match the configuration to the material
Microsoft's high-fidelity music-mode reference describes a mode intended for music content. It recommends turning that mode off for regular calls and meetings. It also identifies availability limits, including live events and use with spatial audio.
These instructions address different intended tasks. Decide which one you are investigating before applying a list of toggles. If you need an instrument or other non-speech source preserved, inspect all processing in that route and test the wanted material with an informed receiver. Neither maker's documentation establishes your own successful result.
For a default-input mismatch, the Mac default versus app selection guide covers another checkpoint. Changing the system default does not document the physical microphone Krisp currently uses or the input Teams displays.
Compare one deliberate choice
Choose a non-critical session and use a non-sensitive phrase or sound you have permission to transmit. Record the earlier selections and the settings to restore. Keep the source position, passage and receiving endpoint consistent while changing one relevant choice.
Ask the receiver to describe the wanted content and anything missing. A local level meter is a local activity observation. Keep it separate from what reached the meeting. If the result is uncertain, restore the intended arrangement and retain the unresolved finding rather than changing every stage together.
For support, provide the app versions, named sources, processing states and one comparison result. Keep account identifiers and private meeting links out of the initial note.
Further reading: a speech-enhancement research example
Jean-Marc Valin's 2018 RNNoise paper, version 3 describes a speech-enhancement approach and its speech-and-noise training material. It is useful context for asking what a processing method is designed to preserve. It does not identify the algorithms used by Teams or Krisp or establish their behaviour with your music.
The cited manufacturer guidance and the paper's relevant method sections were read for this guide. No Teams or Krisp account, audio path, setting, call, recording or listening experiment was operated. The record is blank and provides no compatibility or transmitted-audio result.
The cited reference passages were reviewed on 13 November 2026.