Why your audio is quiet on YouTube

You upload a video, play it next to somebody else's, and yours is the quiet one. So you go back, push the levels up, and export again. It arrives at exactly the level it arrived at before.

Every large platform measures how loud your file is and turns it towards a target of its own. You cannot export your way past that, and pushing harder costs you something on the way. What you can change is how much of your level is voice, which is what actually decides whether you sound loud once the platform has finished with you.

Why is your video quieter than everybody else's?

Not because of your export settings, your microphone, or a switch you have failed to find. Everybody's file gets measured, including the ones you are comparing yourself against.

What differs between two files arriving at the same measured level is what is inside them, and that is decided in the mix rather than at the end.

What does the platform do to your file?

It performs loudness normalisation, which is worth spelling out because the name hides how simple it is. The platform measures how loud your file is across the whole of its duration, rather than at its single loudest instant, using the same algorithm the industry measures with, ITU-R BS.1770. Then it turns your file down if it is louder than a target of its own. That target is not a single fixed number for everyone: some destinations publish theirs and some leave you to measure it, and it can differ between platforms and change over time. Because the platform re-levels your file on playback anyway, it is a figure to master towards with headroom rather than one to land on exactly, so what stays fixed is the measurement and the method, not the destination.

The reason is the viewer. Nobody should have to reach for the volume control between one video and the next, and before normalisation everybody did.

Note what it does not do at the upload. It turns loud files down and does not lift quiet ones, so as far as the platform's own levelling goes, a file that arrives quiet arrives quiet. The viewer's player is a separate matter: on YouTube, Stable Volume is on unless the viewer turns it off, and it evens out the quiet passages after the fact. So a quiet master is not simply preserved, it is either left low by the platform or smoothed by something you do not control, and either way that is the argument for arriving at the right level rather than being corrected into it. It is also why the fix is never at the export stage.

Why does making it louder not make it louder?

Because the measurement is an average across the whole file. If you squash the audio so it is loud all the way through, the average goes up, so the platform pulls it down further. The two files meet in exactly the same place.

The difference is what each of them spent getting there. One arrives with its quiet moments and its loud moments intact. The other arrives with everything at one height, having thrown away the differences that made it worth listening to.

Making it louder does not make it louder. It makes it flatter.

What do you lose when you push?

Dynamics, meaning the difference between the quiet moments and the loud ones. That difference is what makes a sentence land, what makes a demonstration sound like it is happening in front of you, and what stops half an hour of talking becoming a wall the viewer stops hearing. Flat audio is tiring, and tiredness never arrives as a complaint about your audio. It arrives as people leaving.

There is a second cost, and it is the one people notice first without understanding it. Heavy squashing lifts everything, including the parts you did not mean to lift. Your room, your fan and your breath all come up to meet the voice, so a push for loudness quietly undoes the cleaning you did earlier in the day.

How do you check what actually happened?

Judge it after the upload, never before.

Upload it unlisted, watch it on a phone, and play it back to back with a video you think sounds right. Then go and change the mix rather than the export. Almost everybody makes this judgement in the editor, which is the one place in the chain where the platform's decision has not happened yet, so the editor is showing you a version nobody else will ever hear.

Do it once for a video you have already published and you will learn more about your own audio than any meter has taught you.

What to do instead of competing on level

Four things, in the order they pay.

1. Get the voice closer and the background quieter, so more of your level is speech. This is the biggest lever by a long way. A voice recorded close in a quiet room spends nearly all of its level on words. A voice recorded across a room spends much of it on the room, and once the platform sets the overall level, the words land lower. Two files, the same measured loudness, one of them far more present. 2. Even the level out. A file whose level wanders makes a viewer set the volume for the loudest passage, after which everything else is too quiet. Bringing the quiet sentences up to meet the rest does more for how loud you seem than anything added at the end. 3. Leave headroom. A limiter, which is a safety device that stops the loudest peaks going past a set ceiling, is worth having as insurance, and a ceiling of about minus one decibel is a sensible starting point. It catches the one peak nobody predicted, and it makes nothing louder. 4. Stop competing on level. Once every file arrives at the same place, loudness is not the competition. Clarity is.

The expensive advice, which is also the honest one: do not buy a mastering plugin for this. Nothing added at the end of the chain buys back a voice recorded across a room, and most of them flatten the file while failing to.

Where the level stops being an export setting

So the next time yours sounds quiet next to somebody else's, leave the export settings alone. Open the mix, find the three quietest sentences in the video, and bring those up to meet the rest of it.

That is the change the platform will not undo, because you have not asked it to be louder. You have asked it to be more evenly itself, and that is the version a viewer hears as loud.

Sources and evidence

What here is a platform fact and what is our own method. Platform behaviour was checked live on 28 August 2026.

Platform facts. Loudness is measured with ITU-R BS.1770-5 (2023), the industry algorithm. What platforms do with that measurement varies, and it is documented unevenly. Spotify publishes its targets on Spotify for Artists: about -14 LUFS by default, -11 for its loud setting and -19 for quiet, with a true-peak recommendation alongside. YouTube publishes no upload target in that way; its normalisation is established by measurement rather than a stated figure, and its Stable Volume feature is a separate, viewer-side playback control, on by default unless a viewer switches it off, not documentation of how uploads are levelled. So some destinations hand you a number and some leave you to measure it, and because each re-levels on playback anyway, the target is one to master towards with headroom rather than one to land on.

Our method. Mastering for intelligibility and headroom, then reading the actual playback after upload, is how this studio works rather than a rule any platform sets.