Skip to content
fix.fmaudio repair

Audio quality check free, in your browser

Find out what's wrong with a recording before you fix it or send it on. One file gets every check fix.fm runs, in plain words, and nothing leaves your device.

AUDIO QUALITY CHECKCHOOSE A FILE
or drop it here · WAV · MP3 · M4A · MP4

Checking runs on your device. Up to 200 MB and 60 minutes. Format support depends on your browser.

What to expect

Cost and privacy
Free, with no account. Processing runs on your device; your audio is never uploaded.
Use it for
Check one recording for echo, background noise, muffled sound, distortion, low level, one-ear audio, hum, wind, clicks, harsh S sounds and sudden dropouts.
Inputs
WAV, MP3, M4A or MP4 when your browser can decode the audio.
Downloads
A CSV or JSON report with each finding, dropout times and loudness. It contains no audio or original filename.
Limits
Up to 200 MB and 60 minutes, subject to available browser memory. Loudness needs mono or stereo.
Keep in mind
It describes the problems it was built to hear; it does not judge the content or every kind of damage. It changes nothing in your file.

How to check audio quality, in four steps

  1. STEP 01

    Choose a file

    Drop a WAV, MP3, M4A or MP4 recording onto the check, or try the sample. It decodes on your device and nothing is uploaded.

  2. STEP 02

    Read the first sentence

    The top line says how many things need a look. Problems are listed worst first with a plain sentence and the measured number. Checks that came back fine are grouped below.

  3. STEP 03

    Open the fixer for each problem

    Each problem links to the fixer for it. The fixer runs the same check, repairs that one problem, and lets you compare before and after. Dropouts need a retake or a cut in the editor.

  4. STEP 04

    Save the report

    Download a CSV or JSON report with every finding, the dropout times and the loudness readings. It holds no audio and no filename, so you can share it with an editor or client.

What the check listens for

The check runs the same eight tests each fix.fm fixer runs before it repairs anything: room echo, background noise, muffled sound, clipping, low level, audio in one ear, mains hum and wind. Each result starts with a sentence in plain words, then the number it was based on. A check that can't judge your file says so instead of passing it.

It also looks for dropouts: moments where the sound cuts to digital silence for between 3 ms and a second, with sound on both sides. That is what a call breaking up or a recorder skipping leaves behind. Pauses with room tone, fades and silence at the start or end are not counted, and a gap only counts when every channel drops at once.

Loudness is measured with the ITU-R BS.1770 method used by podcast and streaming apps, with an estimated true peak. Everything runs in a Worker on your device, so the file is never uploaded. The thresholds and the samples we test against are published on the how we test page.

Questions

How do I check the quality of an audio file?

Choose the file here. In a few seconds you get a plain sentence for each of nine checks: echo, background noise, muffled sound, distortion, low level, audio in one ear, hum, wind and dropouts, plus loudness and true peak. Nothing is uploaded.

Does it change my recording?

No. The check only reads a decoded copy. To repair something it found, open the linked fixer, which works on its own copy and gives you a new download.

What counts as a dropout?

A stretch of 3 ms to one second where every channel falls to near digital silence, below -90 dBFS, while the audio on both sides is clearly there. Normal pauses keep some room sound and are not counted. Silence at the very start or end is ignored.

Can it tell me if my podcast is ready to publish?

It tells you whether the common problems are present and how loud the file is. It does not check content, edits or a specific platform's delivery rules. Compare the loudness with your destination's guidance, and listen to the finished file.

Why does a check say it couldn't check something?

Some checks need speech or a certain length to judge. For example, echo is hard to measure under heavy noise. The check says so rather than calling the file fine.

What is in the report?

Each check's level, sentence and measure, a link to the fixer, the dropout count and start times, and the loudness readings with units. Missing readings are blank in CSV and null in JSON. No audio, waveform or filename is included.