Why your track is quieter than everyone else's, and what LUFS means

You export a podcast episode, put it next to a professionally produced one, and yours is noticeably quieter — even though the waveform looks like it is touching the ceiling. The reason is that you and the platform are measuring two entirely different things.

Peak is not loudness

The number most audio software shows you by default is peak level: the height of the single tallest sample in the file. Digital audio cannot exceed 0 dBFS, so "normalise to -1 dB" means "scale everything until the loudest instant sits just under the ceiling".

The problem is that a single instant tells you nothing about how loud something sounds. A recording of someone speaking quietly with one door slam in it will peak at 0 dBFS because of the slam, and normalising it does nothing at all — it is already at the ceiling. Meanwhile a dense, consistently loud track with the same peak sounds several times louder, because loudness is about energy over time, not the tallest spike.

This is why "my waveform is huge but it still sounds quiet" is such a common and such a correct observation.

What LUFS measures

LUFS — Loudness Units relative to Full Scale — is the standardised attempt to measure loudness the way an ear experiences it. It comes from broadcast standard ITU-R BS.1770, and it does three things peak metering does not:

• It averages over time rather than reading an instant, so one door slam does not define the file.
• It weights by frequency. Human hearing is most sensitive between roughly 2 and 5 kHz and much less sensitive at the extremes, so the standard applies a filter that mirrors that before measuring. This is why a bass-heavy mix measures lower than it feels on big speakers, and why adding sub-bass does not raise your LUFS much.
• It gates out silence. Passages below a threshold are excluded, so a podcast with long pauses is not dragged down by the pauses.

The result is a single number, in negative decibels, that corresponds closely to how loud a person judges the material to be. It is the number every platform now uses.

Which is why turning it up stopped working

Through the 1990s and 2000s, mastering engineers competed to make records louder — the "loudness war" — by compressing dynamic range until everything sat near the ceiling. Louder genuinely sounded better on the radio and in shops, so it kept escalating, at the cost of every quiet moment in popular music.

Streaming ended it. Spotify, YouTube, Apple Music and the rest all now normalise playback to a target LUFS. Upload something mastered to -6 LUFS and the platform simply turns it down to match everything else. You do not get to be louder. You only get to be more squashed than your neighbours, at the same playback volume — which is a straightforward loss.

Rough targets, integrated over the whole piece:

• Spotify, Amazon Music, YouTube — around -14 LUFS
• Apple Music — around -16 LUFS
• Podcasts — -16 LUFS stereo, -19 mono, and platforms are increasingly strict about it
• Broadcast television (EBU R128) — -23 LUFS, and this one is a legal requirement in several countries

So why is your file quiet?

Usually one of four things.

It was recorded quietly. Gain set too low at the source. This is the honest answer most of the time, and it is also the only one with a real cost: amplifying a quiet recording amplifies its noise floor along with it. Nothing downstream fixes that; it has to be fixed at the microphone.

Wide dynamic range. The loud parts are fine and the quiet parts are far below them, so the average — the LUFS figure — is low even though the peaks are not. Common with untreated speech, where a natural voice moves over a 20 dB range across a sentence.

You normalised the peak. As above: this does nothing for perceived loudness, and it is the default in most simple tools.

One transient is holding you back. A single cough, chair scrape or plosive is at the ceiling and everything else is 15 dB below it. Cut or duck that one moment and the whole file can come up.

Fixing it, in the right order

1. Clean up first. Remove hum, hiss and background noise before touching gain, because every subsequent step amplifies whatever is left. Our noise remover handles the common cases.
2. Deal with the outlier peaks. Find the two or three transients that are far above everything else and reduce them individually. This alone often buys 4–6 dB of headroom.
3. Compress moderately. Narrowing the gap between loud and quiet is what actually raises the average. For speech, a ratio around 3:1 with a few dB of gain reduction on the loudest phrases is plenty. Heavy compression on speech sounds like breathing and pumping, and it is more obvious than being slightly quiet.
4. Then raise the level to hit your target. Once the dynamic range is under control, gain does what you expected it to do in the first place. Our volume booster and audio enhancer work through this sequence rather than just multiplying the samples.
5. Leave headroom. Aim for true peaks around -1 dBTP. Lossy encoding can push peaks slightly above where they were in the source — an MP3 or AAC of a file that peaked exactly at 0 will clip on playback — so that last decibel is not wasted.

The mistake worth avoiding

Do not master to -9 LUFS because it sounds more impressive in your editor. Every platform will turn it down to -14, and what the listener hears is your dynamic range gone with nothing gained for it. The quiet parts of music are not wasted space; they are what makes the loud parts land.

Master to the target. If it sounds quiet next to an old reference track in your editor, turn your monitors up — that is what the volume knob is for, and it is what every listener's player is doing anyway.