Is my audio uploaded anywhere?

No. Everything runs inside your browser using the Web Audio API, so the file never leaves your device. You can prove it: load the page, turn off your Wi-Fi, and the tool still works.

What is LUFS, and why not just normalize the peak?

LUFS is loudness as people hear it: the signal is frequency-weighted to match the ear’s sensitivity, averaged over time, and gated so silence between phrases does not count. Peak normalization measures one single sample — the loudest instant in the file — which tells you nothing about how loud the file sounds. A sparse acoustic guitar recording and a dense compressed mix can share an identical peak while one is obviously quieter than the other. If your goal is that several files sound the same, use LUFS; peak normalization is the setting that people reach for and then wonder why it did not work.

What target should I use?

-14 LUFS is what Spotify, YouTube and Amazon normalize to and a safe default for music going online. Apple Music uses -16, which is also the usual podcast figure and what Apple Podcasts expects. Broadcast delivery in Europe is -23 LUFS under EBU R 128, and -24 LKFS in the US under ATSC A/85. If you are matching a set of files for your own use and none of them are going anywhere in particular, the exact number matters far less than using the same one for all of them.

Why did a file stop short of the target?

Because raising it further would have pushed its peaks past the true-peak ceiling. A quiet recording with a few loud transients — a hand clap, a door, a laugh — runs out of headroom long before its average level reaches the target, and the honest options are to accept it slightly quieter or to squash those transients. This tool takes the first option and tells you how far short it stopped. If you need the target met exactly, lower it, or raise the ceiling if you have headroom to spare.

Does it compress or limit the audio?

No. Every sample in a file is multiplied by the same number, so the dynamics, the balance and the timbre are all exactly as they were — only the level changes. That is a deliberate limit. Compression and limiting are how you fit a loud target into a peaky recording, and they change how the recording sounds, which is not something a tool called "normalize" should do to your file without being asked.

What is a true peak, and why is the ceiling below zero?

The samples in a digital file are points on a curve, not the curve itself. When it is played back the curve is reconstructed between those points, and it can rise above the highest sample — sometimes by more than a decibel. A file peaking at exactly 0 dBFS can therefore clip on playback, and clip again when an MP3 or AAC encoder reconstructs it slightly differently. Measuring the true peak means reconstructing the signal four times over between samples and reading the maximum, which is what this does. A ceiling of -1 dBTP leaves enough room for that overshoot; broadcast specifications typically require it.

No. Loudness normalization is a published standard — ITU-R BS.1770 — and implementing it exactly is both possible and the entire point: two tools that both follow it will report the same number for the same file, which is what makes the number worth anything. A model could only guess at what the standard defines. The AI on Audio Magic is aimed at problems that have no formula: Whisper for speech recognition, DeepFilterNet3 and RNNoise for pulling a voice out of noise, and Demucs for separating a mix into stems.

Do I need an account?

No. The free tools work straight away with no sign-up, up to 15 minutes of audio per tool per day. An account is only needed if you subscribe to remove the limit.

Yes. Every tool on this page is free to use, with a daily limit on total audio processed. There is no watermark and no trial period.

Normalize Audio Volume to a LUFS Target

Six recordings made on six different days are never the same volume, and riding a fader through all of them is the tedious part of finishing anything. This measures each file and applies the one gain that lands it on a target you set — most usefully an integrated loudness in LUFS, the measure Spotify, YouTube and every broadcaster use, so files matched here are still matched after they are uploaded somewhere else. It is arithmetic, not a model: a measurement and a multiply, exact and identical every time. Nothing is compressed and nothing is limited, so a file comes back sounding like itself at a different level, and a true-peak ceiling stops any of them clipping on the way. Peak and RMS modes are there too, for when a specific number is what you were asked for.

How to Normalize Audio Volume to a LUFS Target

1
Choose what to match
Loudness in LUFS is the one that makes files sound equally loud. Peak and RMS are there for when a specification asks for them by name.
2
Set a target and a ceiling
-14 LUFS matches the streaming services, -16 suits podcasts, -23 is broadcast. The ceiling, usually -1 dBTP, is the level no file will be pushed past.
3
Drop your files in
Each one is measured as it arrives, and its row shows what it read, where it will land, and the gain in between — before you commit to anything.
4
Check one by ear, then export
Select a file to see its waveform drawn at the level it will leave at, preview it, then normalize the batch and download each as a WAV.

How it works

Loudness is measured to ITU-R BS.1770-4, the recommendation EBU R 128 is built on. The signal passes through the two-stage K-weighting filter — a high shelf standing in for the acoustic effect of a head, then a high-pass that discards the very low end — and the mean square is taken over 400-millisecond blocks overlapping by three quarters. Those blocks are then gated twice: an absolute gate drops digital silence, and a relative gate drops everything more than 10 LU below the rest, which is what stops the pauses in speech from dragging the answer down and is why two files matched this way genuinely sound equally loud. The filter coefficients are derived for your file’s own sample rate rather than assuming 48 kHz, so a 44.1 kHz file measures correctly. Peaks are read as true peaks: the signal is reconstructed four times over between its samples with a windowed-sinc interpolator, because a waveform that never touches full scale at any sample can still exceed it in between, which is exactly where a converter or a lossy encoder distorts.

Supported formats

WAV, MP3, M4A, AAC, OGG and FLAC. Files are decoded by your browser, so anything it can play will work here. Output is exported as WAV.

Frequently asked questions

Is my audio uploaded anywhere?: +−Is my audio uploaded anywhere?
No. Everything runs inside your browser using the Web Audio API, so the file never leaves your device. You can prove it: load the page, turn off your Wi-Fi, and the tool still works.
What is LUFS, and why not just normalize the peak?: +−What is LUFS, and why not just normalize the peak?
LUFS is loudness as people hear it: the signal is frequency-weighted to match the ear’s sensitivity, averaged over time, and gated so silence between phrases does not count. Peak normalization measures one single sample — the loudest instant in the file — which tells you nothing about how loud the file sounds. A sparse acoustic guitar recording and a dense compressed mix can share an identical peak while one is obviously quieter than the other. If your goal is that several files sound the same, use LUFS; peak normalization is the setting that people reach for and then wonder why it did not work.
What target should I use?: +−What target should I use?
-14 LUFS is what Spotify, YouTube and Amazon normalize to and a safe default for music going online. Apple Music uses -16, which is also the usual podcast figure and what Apple Podcasts expects. Broadcast delivery in Europe is -23 LUFS under EBU R 128, and -24 LKFS in the US under ATSC A/85. If you are matching a set of files for your own use and none of them are going anywhere in particular, the exact number matters far less than using the same one for all of them.
Why did a file stop short of the target?: +−Why did a file stop short of the target?
Because raising it further would have pushed its peaks past the true-peak ceiling. A quiet recording with a few loud transients — a hand clap, a door, a laugh — runs out of headroom long before its average level reaches the target, and the honest options are to accept it slightly quieter or to squash those transients. This tool takes the first option and tells you how far short it stopped. If you need the target met exactly, lower it, or raise the ceiling if you have headroom to spare.
Does it compress or limit the audio?: +−Does it compress or limit the audio?
No. Every sample in a file is multiplied by the same number, so the dynamics, the balance and the timbre are all exactly as they were — only the level changes. That is a deliberate limit. Compression and limiting are how you fit a loud target into a peaky recording, and they change how the recording sounds, which is not something a tool called "normalize" should do to your file without being asked.
What is a true peak, and why is the ceiling below zero?: +−What is a true peak, and why is the ceiling below zero?
The samples in a digital file are points on a curve, not the curve itself. When it is played back the curve is reconstructed between those points, and it can rise above the highest sample — sometimes by more than a decibel. A file peaking at exactly 0 dBFS can therefore clip on playback, and clip again when an MP3 or AAC encoder reconstructs it slightly differently. Measuring the true peak means reconstructing the signal four times over between samples and reading the maximum, which is what this does. A ceiling of -1 dBTP leaves enough room for that overshoot; broadcast specifications typically require it.
Does this use AI?: +−Does this use AI?
No. Loudness normalization is a published standard — ITU-R BS.1770 — and implementing it exactly is both possible and the entire point: two tools that both follow it will report the same number for the same file, which is what makes the number worth anything. A model could only guess at what the standard defines. The AI on Audio Magic is aimed at problems that have no formula: Whisper for speech recognition, DeepFilterNet3 and RNNoise for pulling a voice out of noise, and Demucs for separating a mix into stems.
Do I need an account?: +−Do I need an account?
No. The free tools work straight away with no sign-up, up to 15 minutes of audio per tool per day. An account is only needed if you subscribe to remove the limit.
Is it free?: +−Is it free?
Yes. Every tool on this page is free to use, with a daily limit on total audio processed. There is no watermark and no trial period.

Normalize Audio Volume to a LUFS Target

How to Normalize Audio Volume to a LUFS Target

Choose what to match

Set a target and a ceiling

Drop your files in

Check one by ear, then export

How it works

Supported formats

Frequently asked questions

Other free audio tools