Krisp Review

Reviewed by Tom · Founder & lead reviewer

What it fixes

Every remote worker owns the moment: the courier rings, the dog answers, and the client hears all of it. Hardware solutions cost hundreds and still lose to a leaf blower, because a directional microphone cannot distinguish between the sound you want and the sound that happens to arrive from the same direction.

Krisp fixes it in software by becoming a virtual microphone. Calling applications see a device called Krisp; Krisp hears your actual room; only your voice survives the pass. Because the interception happens at the operating-system level rather than inside any one application, a single setup covers every call you will take - meeting platforms, softphones, browser calls, recording tools.

The bidirectional part matters just as much and gets less attention. Noise from the other participants is cleaned before it reaches your ears, which changes the character of long calls with colleagues who work from call centres, cafes or kitchens. Listening fatigue on those calls is largely a noise problem, and removing it is the difference between an hour that drains you and an hour that does not.

Who it is for

The clearest case is anyone whose calls happen somewhere imperfect: a home with children or animals, a shared flat, a cafe, a coworking space, a car. For those people the product is not a convenience, it is the difference between sounding professional and sounding chaotic.

The second clear case is teams whose job is talking - support, sales, recruitment, customer success. The reason is economic rather than acoustic. The cost per seat is small, the variance in people's home acoustics is enormous, and standardising removes an entire class of problem that otherwise gets solved badly, one person at a time, with headset purchases that do not work.

The third case is the one people discover rather than plan: anyone who records. Interviews, podcast guests, internal training, sales call reviews. A clean input is worth more than any amount of cleanup afterwards.

It is not for the person with a treated room and a broadcast microphone. If that is you, the tool solves a problem you have already solved with hardware.

How the pricing is structured

Our entry records a free tier capped at sixty minutes of noise removal per day, with Pro from roughly eight dollars a month. We could not confirm the current figures against the live pricing page for this update, so open it before committing - but the structure is the part worth understanding, and the structure has stayed stable.

The free tier is a daily minute cap rather than a feature cripple, which is an unusually honest design. You get the real product, you simply run out of it. For someone taking a couple of calls a week that is a permanent free tool. For anyone doing remote work properly it is a trial that expires every afternoon, and the upgrade stops feeling optional inside the first week.

What pushes you up a tier beyond the minute cap: the meeting assistant allowances, team administration and centralised billing, and whatever the current plan defines as enterprise controls. Check the current plan page for which of the assistant features sit on which tier, because that boundary has moved as the assistant has grown.

Day-to-day use

Krisp is built to disappear after installation. It sits between the microphone and the call, so kitchen clatter, cafe noise and the mechanical keyboard you refuse to give up are filtered before colleagues hear them. The processing is designed to preserve accent and tone rather than produce the underwater artefact that cheaper filters are known for, which is the single most important quality distinction in this category.

The meeting assistant runs in parallel: automatic transcripts and summaries that turn who-said-they-would-do-what into a search rather than an argument. This started as an add-on and has grown into a genuine second product, to the point where it quietly replaces a separate notetaker subscription for a lot of users.

The trade-off is processor load, because the suppression runs locally rather than in a data centre. On an older laptop during a large gallery-view call, the fans make themselves known. That is the cost of the privacy property described below, and for most people it is the right trade - but it is a real one, and it is worse on exactly the machines least able to absorb it.

Where the audio goes, and the privacy question

This is the question to ask of any tool that sits in the path of your conversations, and here the answer is unusually good: the noise removal is processed on the device. Your audio does not travel to a server to be cleaned.

That property is worth more than it sounds. A noise filter is in the path of every sensitive call you take - client calls, interviews, medical appointments, legal conversations - and a cloud-processed filter means all of that passing through a third party by design. Local processing removes the question rather than answering it.

The nuance is that the meeting assistant is a different feature with different mechanics. Transcription and summarisation of a meeting is not the same operation as filtering a microphone stream, and if you are evaluating this for a regulated environment, read the vendor's own documentation on where transcript data lives, how long it is retained, and who in your organisation can reach it. Do not let the on-device property of one feature stand in for the whole product.

One more practical point: recording and transcribing a conversation has consent rules that vary by jurisdiction and are not the tool's responsibility. If you turn the assistant on, tell people.

Setup and how it behaves with other tools

Installation is a driver install and then selecting Krisp as the microphone in whatever application you are calling from. That is the whole configuration for most people, and the reason ease is the highest of the four scores in our rating.

The known friction is that per-application microphone settings are per application. Each calling tool keeps its own device selection, so the first week involves pointing each one at Krisp as you use it. Applications that are strict about audio devices, or that update and reset their selection, occasionally need pointing again.

The second thing to know is that stacking noise suppression makes things worse, not better. Most meeting platforms now ship their own filter, and running both means two systems processing the same stream and each reacting to the other's output. Turn the platform's own suppression off and let one tool do the job. If a voice sounds strange after setup, this is the first thing to check.

The third: a virtual device in the audio path is one more component that can be the cause when audio breaks. Knowing it is there saves twenty minutes of confusion the first time something goes wrong.

Limits and the edges where it struggles

Extreme noise is the honest limit. Loud music with vocals, shouting nearby, sustained mechanical noise at close range - these push the filter to where voices go slightly robotic at the edges. The failure is graceful rather than catastrophic, but it is audible, and no software filter escapes it because the information simply is not there to recover.

The processor tax is the second limit, and it is worst in the combination of an old machine and a large call - which is to say, the combination many people actually have.

The third is the free tier's daily cap, which is not really a flaw so much as a clearly signposted boundary. It does mean the free plan cannot be a long-term answer for full-time remote work.

A fourth, more subtle one: the product is so unobtrusive that people forget which microphone they are using, and then take a critical call on a different device wondering why it sounds worse. Check the device selection before anything that matters.

Alternatives and adjacent tools

The nearest alternatives are not other noise filters for most buyers - they are the two other ways of solving the same problem. One is hardware: a better microphone in a better room, which works and costs considerably more. The other is the noise suppression already built into your meeting platform, which is free, improves steadily, and is genuinely adequate for mild noise in a quiet home. The case for a dedicated tool strengthens as your environment gets worse, and it is strongest for the bidirectional cleaning, which platform filters do not do for you.

On the assistant side, the overlap is with dedicated meeting tools, and those are deeper if notes are the main thing you want: see the Otter.ai review, the Fireflies.ai review, the Otter vs Fireflies comparison and the best AI meeting tools guide.

The sensible arrangement for most teams: this tool for the microphone path, and a dedicated meeting tool only if the notes matter more than the audio. Buying both for the same job is a common and avoidable duplication.

Common questions

Does it work with every calling application? It presents itself as a microphone at the operating-system level, so anything that lets you choose a microphone can use it. Each application needs pointing at it once.

Does my audio leave my computer? The noise removal is processed on the device. The meeting assistant is a separate feature with different mechanics - read the vendor documentation if that distinction matters to you.

Should I turn off my meeting platform's own noise suppression? Yes. Running two filters on one stream degrades the result.

Will it fix a bad microphone? It removes noise around your voice. It does not add quality your microphone never captured.

Is the free tier usable long term? Only for light use. Sixty minutes a day is a real boundary for anyone working remotely full time.

Should you buy it?

If you take calls from imperfect places - homes, cafes, coworking spaces, the road - the free tier will make the case inside a week and the upgrade will feel obligatory after. The professionalism gained per dollar here is out of proportion to the price, which is why this is one of the easier recommendations on the site.

Support and sales teams should treat it as standard issue. The variance in people's home acoustics is the problem being solved, and solving it centrally is cheaper and more reliable than solving it with individual headset purchases.

Skip it if you have a treated room and a broadcast microphone, in which case you are the person the rest of us sound like because of tools such as this one. Skip it also if your calls are already quiet on both ends and your platform's built-in filter handles the occasional lapse - that is a genuinely adequate answer for a meaningful number of people, and it is free.

Best for

Remote workers, support teams and anyone whose calls happen in kitchens, cafes and homes with children or dogs.