How we test background noise removers

Last updated:

We make Denoisa, so treat our tests as evidence from a vendor, not an independent review. This page explains the recordings we use, what we measure and where our results stop, so you can check them against your own recordings.

Try it on your recording: free 30-second preview

The recordings we use

Our noise test uses 12-second samples of a narrator from a public-domain LibriVox recording, mixed with real noise recordings released under CC0 on Freesound. We mix speech and noise at a similar loudness, so the noise sits only a few decibels below the voice. That is harsher than most real recordings.

The test page has playable clips, so you can listen rather than rely on our numbers.

What we measure

We measure three things on 50-millisecond slices of each file: the noise floor (the quietest 10% of slices, mostly the gaps between words), the speech level (the loudest 5%), and how much speech energy sits above 4 kHz. The last number matters because consonants like s, t and f live there. A tool can score well on noise and still leave a voice sounding muffled.

For video, we check whether the picture really stays the same. We ran an H.264 MP4 and an H.265 MOV through Standard and compared the video data in each output with the original.

What our tests can’t show

Our noise test is small: four samples with one narrator each. The noise was mixed in artificially, and the measurements compare each file with itself rather than with the clean original, so they describe noise and brightness, not overall quality. We have not tested echo, cheap microphones or clipping, and we have not finished testing on phones.

The numbers are a guide to what kinds of noise are hard, not a benchmark score. The fairest test is your own recording: every file gets a free 30-second preview of both modes.

Rules for comparing other tools

When we compare Denoisa with another tool, we use the same recordings for both and note the date of the test. Prices and limits come from the tool’s own website, with the date we checked them, because they change.

We publish the results where another tool does better, and say when we would recommend it instead. We don’t use anyone else’s audio or logos, and we don’t make up reviews, ratings or user numbers.

Corrections

If you find a mistake in a test or a comparison, write to support@denoisa.com. We check the report, correct the page and change its last updated date.

Hear the difference in 30 seconds

Choose a file