Start a project →

Free Audio Dropout and Click Detector: find where the sound breaks

Choose a video or audio file and see the moments where the sound breaks: dropouts where it vanishes for a few milliseconds, clicks and pops at edit points, and cuts that start or stop the sound with no fade. Each one has its time, length and how loud it is, so you can click it and listen. Pauses, plosives and drum hits are left alone. In my tests it found every dropout to silence of 20 ms or longer, but small clicks can slip through, above all in MP4 and MP3 files. Everything runs in your browser: your file never leaves your device.

Updated · Built by Kevin Tabares, long-form YouTube editor at Umbrella Creators.

Find dropouts, clicks and abrupt cuts in a video or audio file

Your file stays on your device. Nothing is uploaded: your browser decodes the audio and checks every millisecond right here, in this tab.

How to use it

  1. Choose the file: a raw recording to check the mic, or your finished export to check the edit (MP4, MOV, WEBM, MP3, WAV or M4A, up to 60 minutes).
  2. Press Find dropouts and clicks. The browser decodes the audio and checks every channel; a long file takes some seconds and the bar shows the progress.
  3. Read the verdict, look at where the marks fall on the strip and click any time to hear 3 seconds around it. Download the CSV, the label file for Audacity or the EDL for your video editor, to fix them.

WAV files from a recorder give the most exact answer. In MP4, M4A and MP3 files the codec smooths small clicks and breaks shorter than about 10 ms, so check the original recording when a short glitch matters.

How I tested it

I put breaks with a known time into two real recordings: a 13 minute documentary narration and the 12 minute audio of an edited talking-head video. Then I ran the files through this page in Chrome, as WAV and as AAC at 192 and 128 kbps (the audio of most MP4 videos), and counted what it found within 10 ms of the right time (5 ms for clicks). For the fade rows I cut 5 more minutes of each recording to silence at 80 places in running speech, once with no fade and then with the fades the fix box below suggests. The untouched recordings, more narrations and 18 music tracks were checked too, because a detector that flags normal sound is useless.

What I put inWAVAAC 192 kbpsAAC 128 kbps
Dropouts of 5 ms (digital silence) in speech4 of 44 of 4 (3 as a click)3 of 4 (3 as a click)
Dropouts of 20, 100 and 500 ms in speech12 of 1212 of 1212 of 12
Sound turned down 40 dB for 50 ms2 of 22 of 22 of 2
Sound cut off mid-word, then 2 s of silence4 of 44 of 44 of 4
Sound cut back in mid-word4 of 44 of 44 of 4
Clicks and jumps in a pause14 of 1413 of 148 of 14
Clicks and jumps inside speech15 of 2410 of 245 of 24
Edits with no crossfade, mid-word (each side)4 of 123 of 121 of 12
Speech cut off with no fade (to 0.2 s or 1.2 s of silence)24 of 2424 of 2424 of 24
Must find nothing: the same cuts with a 20 or 30 ms fade out and fade in67 of 6763 of 6763 of 67
Must find nothing: the same cuts with a 10 ms fade (linear or equal power)43 of 4640 of 4640 of 46
Must find nothing: faded gaps and silences, crossfaded edits, a 15 dB dip12 of 1212 of 1212 of 12

Counts are for the two recordings together. On the untouched files it listed 1 moment in the 13-minute narration, a place where the narration file itself stops sharply, and nothing in the 12-minute video or in 12 raw camera takes. In 3 more narrations it listed nothing in one, 1 Low stop in another, and 14 breaks in a third, at places where a word starts at full level straight out of near silence, or a word is cut off. In 18 music tracks it found nothing in 16, including a percussion track full of drum hits; it listed 9 clicks in a track styled as glitched vinyl and 86 in a synth track whose waveform jumps in steps. A 59.5-minute stereo AAC file took about 7 seconds in Chrome and all 5 dropouts I put in it were found.

FAQ

What counts as a dropout or an abrupt cut, and what does not?
A dropout is a moment where the sound vanishes or falls by a lot in the middle of continuous sound: a wireless mic losing signal, a bad cut, a glitch in a recorder. The tool lists one when the level falls 33 dB or more within about 1 ms, from steady sound, and stays at least 35 dB down for 3 ms to 1 second; when the sound stays away for more than a second, it is an abrupt stop. A pause between sentences is different: the word fades out first, so if the sound was already 12 dB quieter than the 100 ms before, nothing is listed. A fade is left out too: when the last 1.5 ms before the break are 18 dB or more under the loudest moment of the 15 ms before it (and, for a dropout, the sound also fades back in). In my tests, cuts in running speech with a fade of 20 ms or more were left out 211 times out of 221, and with a 10 ms fade 133 times out of 154. A fade on a word that started less than 50 ms before the cut was still listed 16 times out of 20, because the word is chopped either way: move that cut into the pause.
Why are plosives, drum hits and the silences between sentences not flagged?
Because they behave differently from a break. A click is a sample jump that stands alone: the tool needs it to be 24 times stronger than the sharp detail in the 2 ms before it and also in the 2 ms after it. A drum hit or a "p" goes on after its attack, so the 2 ms after it are loud too. An abrupt start needs 30 ms of silence before it and a waveform that rises in 0.35 ms or less; in the narrations I measured, natural starts after silence took 0.45 ms or more, while sound cut in mid-word took 0.13 to 0.27 ms. Silences between sentences start with the word fading out, so they are not dropouts. In 18 music tracks, a percussion track full of drum hits among them, it listed nothing in 16.
What can it miss?
Short glitches, above all in compressed files. In my tests it found every dropout to silence of 20 ms or longer, in WAV, AAC and MP3, and 137 of 138 cuts with no fade from speech to silence; a cut to room tone instead of silence was found only 8 times out of 20. Clicks are another story: a click or a jump inside loud speech was found 35 times out of 54 in WAV but only 44 out of 118 after AAC or MP3, and an edit with no crossfade 15 times out of 31 in WAV and 22 out of 69 after AAC or MP3. A 5 ms dropout was found in all 13 WAV cases but in 14 of 29 compressed ones, often listed as a click. So a clean result on an MP4 does not prove there are no clicks: check a WAV export when it matters. An abrupt start is only listed when the sound cuts in after near-digital silence (under -80 dB) and comes back within 20 dB of the loud parts; a cut back in on room tone or on a quiet syllable is not. Sounds that are abrupt on purpose, like a stutter edit, vinyl crackle, a synth with hard edges or a hard music stop, are listed, because they are abrupt. The limit is 60 minutes per file, and long files need a lot of memory: Chrome on my Mac reached about 9 GB in total while it checked a 59 minute stereo file, and still held almost 8 GB a few seconds later. On a phone, or a computer with 8 GB of memory, check long files in parts and close the tab afterwards. In Safari, times from an MP3 came out 12 ms late in my test (Chrome was exact), and Safari could not read the audio of a 227 MB MP4 video that Chrome read in full: export the audio as M4A or MP3, or use Chrome.
What do High, Medium and Low mean?
They tell you how loud the sound was at the break, compared with the loud parts of your file (the level that 95% of the file stays under). High: within 12 dB of the loud parts, you will hear it on any speaker. Medium: 12 to 24 dB lower, normal speech or music level. Low: quieter than that, at the tail of a word or in a quiet passage, often only audible on headphones. For a click, the level is the size of the jump itself.
How do I open the breaks in Audacity or DaVinci Resolve?
For Audacity, download the label file and choose File, Import, Labels. Each break becomes a label at its time (a dropout covers its whole length). The file is the tab separated text format Audacity describes in its manual: start time and end time in seconds, then the label text (Audacity manual, Importing and Exporting Labels, read October 21, 2026). For a video editor, choose the frame rate of your timeline and download the EDL: it is your file cut at every break (a dropout at both ends), so each break becomes an edit point, with a comment that names it. In DaVinci Resolve, add your file to the Media Pool and choose File, Import AAF, EDL, XML, as the DaVinci Resolve 20 Reference Manual describes in its chapter on conforming EDL files (read October 21, 2026). The EDL holds up to 998 breaks, the High ones first. The CSV has every time in seconds and as hours:minutes:seconds.
Does it upload my file?
No. Your browser decodes the audio and checks it on your own device; the file never leaves this tab, and nothing is downloaded either.

Who made this

I'm Kevin Tabares. I've edited long-form YouTube videos since 2015 (1,277 videos and 152.2M+ views tracked on YT Jobs) for channels like Rex & Alexa, Puff and Mud. A click on a jump cut or a word chopped by a hard cut is easy to miss in a long timeline and easy to hear on a phone, so I built this to list them in seconds, without uploading anything. If you want an editor for your channel, get in touch.

More free tools

An audio clipping detector for distorted peaks, a silence detector for long pauses and a background noise checker for hum and hiss, all free and without signup.

See all tools →