Record
Hold your phone near the noise for 5 to 10 seconds. Phone-call noise filtering is switched off, so the sound comes through as it is.
Free · In your browser
What’s that noise your car is making?
Hold your phone near the sound and record 5 to 10 seconds. Revnosis compares it with known car sounds and tells you which one it’s closest to. When it doesn’t know, it says “Not sure.”
Live demo: record or choose audio and see its analysis. The reference library is empty, so results currently say “Not sure.” Community sharing is unavailable. No account; your recording stays on your device.
How it works
Three steps, about fifteen seconds.
Compare
A sound model running in your browser turns the clip into a fingerprint and measures how close it is to every reference sound.
Get an honest answer
You see the top three matches with their scores. If nothing is close enough, it says “Not sure” instead of guessing.
Built to be honest
A car noise is easy to get wrong. Revnosis is designed around that.
“Not sure” is an answer
Every score is shown, and so is the cut-off. Unfamiliar sounds get flagged rather than forced into a match.
Your audio stays with you
Matching runs on your device. Nothing is uploaded unless you choose to share a clip.
Not a diagnosis
It compares sounds. Anything involving brakes, steering, wheels or tires gets a clear “get this checked before driving far.”
Sounds it’s learning
The reference library is empty. These are the sounds planned for it; until confirmed reference clips are added, Revnosis says “Not sure.”
Healthy idle
Cold start
Rev
Misfire
Squealing belt
Ticking lifters
Rod knock
Grinding brakes
Wheel bearing
Power steering whine
Tire noise
Red means a safety sound: get it checked before driving far.
About
Revnosis is “Shazam for car noises”: record a noise and find out what it’s closest to.
The idea
Most people can hear that something is wrong with their car but can’t name it. Mechanics can. Revnosis puts the two together: owners share clips, hobbyists and mechanics vote on what they hear, and when a shop finds the real cause, that clip becomes a confirmed reference.
The community side is still being built. Today the demo records and analyses audio on your device. Matching needs reference clips, which have not been added yet.
Made by CKAI.
Built with
- YAMNet by Google, a sound model that turns audio into a fingerprint. Apache License 2.0.
- TensorFlow.js by Google, which runs that model in your browser. Apache License 2.0.
- Revnosis app code. Apache License 2.0.