Can anyone share an honest Krisp AI user review?

I’ve been testing Krisp AI for background noise removal on calls, but I’m getting mixed results with audio quality and performance. Sometimes it works great, other times it cuts out parts of my voice or lags. I need help understanding if I’m using the settings wrong, if my system is the issue, or if there are better alternatives. Can anyone share real-world Krisp AI user reviews, tips, or recommended configurations for stable call quality?

I’ve used Krisp on and off for about a year for Zoom, Meet, and Discord. Mixed bag is a good way to describe it.

What works well for me:

  1. Constant low noise

    • Keyboard clacks, fan noise, HVAC, people walking in the hallway.
    • It filters those pretty cleanly.
    • Voice still sounds natural enough if your mic is decent.
  2. Short, sharp background sounds

    • Dog bark in another room, someone dropping something.
    • Those get muted most of the time.
    • Better than the built in noise suppression in Zoom or Meet in my tests.

Where it goes wrong:

  1. Aggressive voice cutting

    • If you speak softly or trail off at the end of sentences, Krisp sometimes decides your voice is noise.
    • Happens more when:
      • Input gain is low.
      • You have a lot of background chatter on calls.
      • You talk fast and pause often.
    • On my calls, people told me parts of words went missing, like “impor–ant” instead of “important.”
  2. Lag and CPU issues

    • On a laptop with weak CPU, Krisp ate 10 to 20 percent CPU during calls.
    • When I had Chrome, Slack, VS Code, and a couple of VMs open, I got:
      • Small delay in my audio.
      • Occasional robotic voice artifacts.
    • On a desktop with a Ryzen 5 and 16 GB RAM, no lag problems.
  3. Music or complex sound behind you

    • If someone is talking behind you while you talk, Krisp struggles.
    • It tries to guess which voice to keep and sometimes damages your own voice.
    • Same thing with TV or music in the same room.

What helped a lot:

  1. Mic positioning and gain

    • Use a closer mic if possible, like a headset or a dynamic mic near your mouth.
    • Turn up input gain a bit so your voice is clearly louder than noise.
    • In Krisp, try “Low” or “Medium” suppression first. “High” made my voice choppy more often.
  2. Use Krisp only once in the chain

    • Turn off noise suppression in Zoom, Teams, Discord if Krisp is on.
    • Double suppression made my audio thin and glitchy.
    • Pick one app to do the processing.
  3. Hardware checks

    • On Windows, check Task Manager during a call. If CPU is over 80 percent, Krisp has a harder time.
    • Close Chrome tabs that stream video or run heavy scripts.
    • Use wired headphones. Bluetooth adds its own delay and compression.

Quick comparison to other options I tried:

  • Zoom “High” noise suppression:
    Better echo handling, worse background voices filtering, more robotic sound.
  • NVIDIA RTX Voice / Broadcast:
    Much stronger noise removal, but needs NVIDIA GPU. Less voice cutting for me.
  • RNNoise based tools (like some OBS plugins):
    Similar behavior, but slightly more consistent at low suppression levels.

Concrete suggestions for your case:

  1. If you get cutouts:
    • Drop Krisp to Medium.
    • Raise your mic input level in OS.
    • Speak a bit closer and more consistently.
  2. If you get lag:
    • Check CPU load during calls.
    • Close other heavy apps.
    • Try Krisp at a lower quality setting if available.
  3. If nothing fixes it:
    • Use Krisp only when noise is bad, not for every single call.
    • For quiet environments, use your raw mic audio, or only the built in suppression of Zoom or Teams.

Short version:
Krisp helps a lot when noise is constant and your mic setup is decent. It hurts when CPU is stressed, suppression is set to High, or your speaking volume is low and inconsistent. For some setups it is “set and forget”. For others it needs tuning or only situational use.

I’m in the same “sometimes magic, sometimes disaster” camp with Krisp, but my experience is a bit different from what @reveurdenuit described.

Where I actually like Krisp:

  • Group calls where everyone else is noisy
    I use it mostly to clean incoming audio, not my own. That’s the underrated use case. Colleagues with leaf blowers, kids, clacky keyboards… Krisp on the input side was meh for me, but on the output side it made chaotic meetings survivable.
  • Sudden loud sounds close to the mic
    It handled my own typing and accidental desk bumps decently, without wrecking my voice too often.

Where it drove me nuts:

  • Non‑standard speaking patterns
    If you’re monotone, speak quietly, or have an accent, it can misclassify parts of your voice as noise more often. In my case, when I switched to a more “presenter” style (slightly louder, more dynamic), cutouts reduced a lot even with the same hardware and settings. So it’s not just mic gain like @reveurdenuit said, but also how you speak. That part annoyed me, honestly. A tool that wants you to change your speaking style feels backwards.

  • Long meetings
    After 45–60 minutes on my older laptop, I’d see a slow creep in artifacts and delay, like the model was barely keeping up over time. CPU wasn’t pegged, but the combination of video calls, screen sharing, and Krisp consistently produced this “rubber band” feeling in my audio. Restarting Krisp between meetings helped more than any setting tweak.

Stuff that helped that I don’t see mentioned much:

  1. Per‑app choice instead of global
    Instead of routing everything through Krisp at the OS level, I used it only with the one app that really benefitted (Meet in my case). Other apps (Discord, Slack huddles) stayed on their native noise suppression. That avoided some weird routing issues and random glitches I was getting when multiple apps tried to use the virtual device at once.

  2. Sample rate & audio chain sanity check
    Mismatched sample rates caused more weirdness than CPU for me. On Windows in particular:

    • Set your mic to 48 kHz in the OS.
    • Make sure your conferencing app is also using 48 kHz when possible.
      When I had 44.1 kHz in one place and 48 kHz in another, Krisp artifacts got noticeably worse. No one tells you this, but it matters with realtime processing.
  3. Don’t pair it with “smart” mics
    USB mics or headsets that already have onboard DSP (AGC, compression, noise gate) sometimes fought with Krisp. My gaming headset with its “AI noise cancelation” + Krisp was a mess. A very plain dynamic mic into a simple interface behaved way better. If your mic has its own “enhancement” app, try disabling it entirely and let only Krisp handle things.

  4. Consider flipping your usage pattern
    For you specifically:

    • If people complain they can’t hear parts of your voice, use Krisp primarily on incoming audio and leave your outgoing mostly clean or on light built‑in suppression.
    • You get the benefit of sanity on your side without risking that weird “impor–ant” effect on their side.

My honest verdict:

  • In a noisy shared environment, Krisp is worth keeping installed and using strategically, not as a default “always on” thing.
  • In a quiet or moderately noisy place, I actually prefer decent mic technique + the built in Zoom/Teams suppression. Less risk, less fiddling.
  • If it’s cutting your voice and lagging even after some basic tweaking, I’d treat Krisp as a “break glass when construction starts next door” tool instead of part of your everyday stack.