I work as a freelance UX consultant, and for the past year I’ve been running my business from a one-bedroom apartment that sits directly above a café with an espresso machine that sounds like a small jet engine. Add in the construction crew that started renovating the building next door in March, my neighbor’s enthusiastic Pomeranian, and the fact that I take roughly 15 to 20 client calls a week, and you can probably guess where this is going. Every meeting started with me apologizing for the background noise, fiddling with my mic settings, or awkwardly muting myself mid-sentence whenever something rumbled outside.
I’d tried the obvious fixes. I bought a Blue Yeti, then returned it for a Shure MV7 thinking the dynamic pickup pattern would solve everything. It didn’t. I tried Zoom’s built-in noise suppression on the “high” setting, which mostly just made my voice sound like it was being transmitted from inside a tin can. I hung moving blankets behind my desk, which my partner found extremely charming. I even considered renting a coworking desk just for calls, but the math didn’t work out for someone who spends most of the day in deep work mode and only needs quiet for specific windows.
The breaking point came during a discovery call with a fintech client I’d been chasing for two months. Right as the founder was explaining their roadmap, a jackhammer started up across the street. I had to ask him to repeat himself three times. That night I went searching, ended up reading a krisp ai review on a productivity blog written by a remote sales manager, and figured I had nothing to lose by trying it. The pitch was simple: AI that removes background noise from both sides of a call, works with any conferencing app, and runs as a virtual microphone.
I signed up for Krisp the next morning before my 9am standup. Setup took maybe four minutes. You install the desktop app, it asks for microphone permissions, and then you go into Zoom or Google Meet or whatever you use and select “Krisp Microphone” as your input device. That’s it. No configuration sliders, no calibration wizard asking me to read sentences aloud, no profile creation. I half expected to spend my lunch break troubleshooting, but it just worked.
My first real test was that standup. I deliberately didn’t warn anyone I was trying something new because I wanted honest reactions. About ten minutes in, my project manager asked if I’d moved offices. I told her no, same apartment, and she said my audio sounded “weirdly clean.” Meanwhile, on my end, the café noise downstairs was at its usual mid-morning peak and I could hear it through my open window. But on the call, gone. The Pomeranian barked twice during the meeting and nobody on the call heard a thing. I did notice that when I tapped my desk with a pen, that sound came through, so it’s not removing every transient noise, just the steady-state stuff and unexpected bursts.
What surprised me was the two-way feature. Half my clients work from home too, and I’d gotten used to hearing kids, dishwashers, traffic, and the occasional leaf blower coming through their end. Turning on noise cancellation for incoming audio meant I could actually focus on what people were saying instead of mentally filtering out chaos. One client has a parrot. I had no idea until she mentioned it on our fourth call, because Krisp had been quietly removing it the whole time.
These days I leave it running by default. My workflow is basically: open laptop, Krisp launches on startup, every call I take regardless of platform automatically gets cleaned up. I’ve stopped apologizing for noise. I’ve stopped tensing up when I hear the espresso machine fire up downstairs at 2pm. The time savings aren’t dramatic in any single instance, but across a week of 15 calls, the cognitive load of constantly managing my audio environment has basically disappeared.
It’s not perfect. The free tier gives you 60 minutes of noise cancellation per day, which I burned through in a single morning during my first week, so I ended up on the paid plan at around $8 a month billed annually. Fair, but I wish there was a middle tier for people who don’t need the meeting transcription and AI note-taking features that come bundled in higher plans. I also noticed that on very long calls, two hours plus, my laptop fan spins up noticeably because the noise processing does eat some CPU. Not a problem on my M2 MacBook, but my older Intel machine struggled.
The other quirk: occasionally it’ll clip the very beginning of a word if I start talking right after a loud noise. Maybe once or twice a call. Nobody’s ever mentioned it, but I notice it when I review recordings. And there’s one scenario where I still go back to the old way, which is when I’m recording voiceovers or podcast guest appearances. For those I want the raw audio so I can process it myself in post, because the AI cleanup, while great for live calls, has a slightly compressed quality that’s noticeable in a finished produced track.
I’m planning to test the meeting transcription feature next month when I have a research project with about 12 user interviews scheduled. Right now I pay for Otter separately, and if Krisp can handle that decently I might consolidate. Still figuring out whether the transcription quality holds up for technical UX vocabulary, which is where most automated tools fall apart for me.
