A row of faders on a lit mixing desk, the channel that rides the level of the take
CUTROOM

All tools / Fade Audio

Fade Audio.

Fade in, fade out, on the ends that need it.

→ Set the levelFree, no account, and your video is never uploaded

How it works.

Make a video louder or quieter, and fade the ends.

  1. Step 1

    Drag your file in

    Video or audio. It stays on your device.

  2. Step 2

    Read it, then aim

    The reading tells you where you are.

  3. Step 3

    Save the fixed clip

    You see the level before and after, then download.

In detail

A hard cut into audio reads as a mistake even when the picture is fine, and a cut out on the last word of a sentence is worse. A fade is the fix, and it is two numbers rather than a project file: how long the sound comes up over at the start, how long it goes down over at the end. Set both in half second steps up to ten seconds and the ramps are drawn straight onto the waveform, so you can see how much of the take each one covers before anything runs. The gain and the loudness target are on the same screen if the clip needs both, and the picture is copied across either way.

What it gives you.

  • The same meter the platforms use

    Gated integrated loudness to ITU-R BS.1770, the standard every streaming service normalizes against. Two pre-filters that approximate a head in a sound field, 400 millisecond blocks that overlap by three quarters, and a gate that throws out the pauses so a quiet passage in the middle of a take does not drag the number down. The reading you see is the reading YouTube will take.

  • Normalize to minus 14, minus 16, or minus 23

    Minus 14 LUFS is where YouTube, Spotify, TikTok, and Instagram hold everything they play. Minus 16 is what Apple Podcasts and most podcast players want. Minus 23 is EBU R128, which is broadcast television. Pick one and the gain is arithmetic rather than guesswork, or type your own number if you are delivering to a spec.

  • A peak ceiling that admits when it stopped you

    Turning a clip up moves its peaks up by exactly the same amount, and a sample past 0 dBFS is a click on one player and a crunch on the next. So the gain is capped at whatever keeps the loudest sample under minus 1 dBFS, and when that happens the tool names both numbers: the gain it wanted, the gain it took, and the loudness you will actually land at.

  • Fades in and out

    Linear ramps at either end, zero to ten seconds in half second steps, drawn over the waveform so you can see how much of the take they cover. On a clip shorter than the ramps they multiply into a triangle instead of contradicting each other, which is what an editor would expect.

  • The picture copied, never re-encoded

    Only the sound is decoded and written again. The video packets are copied across byte for byte, so a 4K recording turned up by six decibels comes out at the resolution, the frame rate, and the quality it went in at, and it takes about as long as copying the file would.

  • Works on MP3 and WAV too

    An audio file is not a special case here. Drop an MP3 and an MP3 comes back, drop a WAV and a WAV comes back, at the source channel count and sample rate. The same meter, the same targets, the same ceiling, so a podcast episode and a phone video get the same treatment.

Straight answers.

How do I add a fade in or fade out to audio?

Set the length of each ramp and export. The ramps here are linear, which is what an editor draws and what a listener expects on a tail of a few seconds, and they are applied per sample to the decoded audio rather than as metadata some players would ignore. Both ends are independent, so you can fade in over one second and out over three, and on a clip shorter than the two ramps together they multiply into a triangle rather than fighting each other. The video is copied rather than re-encoded, so adding a fade to the sound of a long recording costs the audio encode and nothing else.

What is LUFS, and why minus 14?

LUFS is Loudness Units relative to Full Scale: a measurement of how loud something sounds rather than how big its numbers are. A quiet solo voice and a dense music bed can peak at exactly the same sample value and be nowhere near as loud as each other, which is why peak meters are useless for this and why the broadcast world agreed on a perceptual standard instead. The meter weights the signal roughly the way an ear does, averages it over the whole programme, and gates out the silences. Minus 14 matters because that is roughly where YouTube, Spotify, TikTok, and Instagram normalize everything they play: upload something louder and it is turned down for you, upload something quieter and on most of them it is left quiet. Landing on minus 14 yourself means the platform does nothing and your clip sits at the same level as everything around it.

Why did it not reach the target I picked?

Because the peak got there first. Loudness and peak are two different measurements, and applying gain moves both of them by the same number of decibels. A clip that averages minus 25 LUFS but has one door slam peaking at minus 2 dBFS cannot be turned up eleven decibels without that slam landing nine decibels past full scale, which is distortion the file cannot store. So the gain is held at whatever keeps the loudest sample under minus 1 dBFS, and the tool tells you the gain it applied and the loudness you will actually land at instead of pretending. If you need to close the rest of the gap, the honest fix is compression or limiting on the original mix rather than more gain here.

Does this touch the video at all?

No. The video packets are copied into the new file without ever being decoded, so the picture in the output is bit for bit the picture in the source: same resolution, same frame rate, same codec, same generation. Only the audio track is decoded, multiplied, and encoded again. That is why the export is fast on a large file, and it is why you can run a clip through here after grading or editing without giving up a generation of quality. The one exception is a file the browser cannot parse at all, such as an AVI, which is converted once with the fallback decoder first, and the tool says so before it starts.

Can I normalize an MP3 or a WAV?

Yes, and it stays what it was. An MP3 comes back an MP3, a WAV comes back a WAV at 16-bit PCM, and an M4A comes back in an MP4 container. The channel count and sample rate are kept, the same BS.1770 meter reads the file, and the same targets and peak ceiling apply, so a podcast episode measured at minus 21 LUFS can be brought to minus 16 for Apple Podcasts in one pass. A lossy file is re-encoded to carry the new levels, because there is no way to change the samples without writing them again, so the export is one generation on from the source. A WAV is not, since PCM is written rather than encoded.

Other tools