Free · no login · nothing uploaded
Audio normalizer.
Level it the way a room hears it.
Normalise a track to a real loudness target rather than to its loudest sample, with a true-peak ceiling that never gets breached. Nothing is compressed, nothing is limited, and nothing is uploaded — it is one gain change, measured properly, and it tells you exactly what it did.
Drop tracks to level
WAV, AIFF, FLAC, MP3 or M4A. Drop a whole folder’s worth — one set of settings is applied to all of them and they come back as a single archive. Nothing is uploaded; the page makes no network request.
Why peak normalising fails
Two tracks at 0 dB are not the same volume.
It is the single most common piece of bad advice in audio, and the reason a normalised folder still needs riding the gain.
Take a stripped-back deep house cut and a peak-time techno roller, normalise both so their loudest sample sits exactly at 0 dBFS, and drop them into the same set. The techno track will be several decibels louder. Nothing went wrong — peak normalising did exactly what it says, which is match one instantaneous sample out of ten million and ignore everything else.
What the ear responds to is average energy over time, weighted towards the frequencies it is most sensitive to. That is what a loudness measurement does, and it is why Spotify, Apple Music, YouTube and every other platform turns your files up or down to a fixed loudness target before anyone hears them. If you normalise a folder to peak and then upload it, the platform undoes your work and does it properly.
So the useful version of this tool targets loudness. Pick where the file is going — a streaming upload, a DJ pool, a club-ready master — and it gets there with one gain change and a ceiling it will not cross. What it will never do is compress the life out of a track to reach a number, because that is a decision about the music and it is not one a converter should be making for you.
Loudness is not the same as level
A peak normaliser finds the single loudest sample in the file and scales everything so that sample sits at your target. That tells you nothing about how loud the track sounds, because a sparse, dynamic arrangement and a wall of limited noise can share the same peak and be eight decibels apart to the ear. Loudness normalisation measures the whole file the way hearing does — frequency-weighted, gated so silence does not drag the average down — and every streaming platform switched to it a decade ago for exactly that reason.
The ceiling is on the true peak, not the samples
A file can contain no sample above full scale and still go over it. The stored samples are points on a waveform, and the waveform between them routinely rises higher than any of them — which is what a converter reconstructs and what a lossy encoder recreates. So the ceiling here is measured at four times the sample rate, and it is a hard constraint: if hitting the loudness target would breach it, the file stops at the ceiling and the tool tells you how far short it stopped.
It will not squash anything to hit a number
Every other normaliser that promises a loudness target reaches it by limiting, which changes the music rather than its level and does so silently. This one applies a single gain and stops. If your track cannot get to -9 LUFS without going over the ceiling, that is a real fact about the track and the answer is a mastering decision, not a checkbox. You get the shortfall in the report and you get to decide.
The targets
Where each number comes from
Pick the destination and the number follows. These are not preferences, they are published platform behaviour.
| Target | Who uses it | What happens if you ignore it |
|---|---|---|
| -14 LUFS | Spotify, Tidal, YouTube, Deezer. | Anything louder is turned down to here on playback. Mastering to -6 buys you no volume at all and costs you every dynamic you crushed to get there. |
| -16 LUFS | Apple Music and Amazon. | Same mechanism, a slightly quieter target. A file mastered for Spotify plays about 2 LU quieter here, which nobody notices. |
| -12 LUFS | A working compromise. | Loud enough to hold up in a set, quiet enough that a platform barely touches it. A reasonable default if the file is going to both. |
| -9 LUFS | Club masters and most DJ pools. | This is roughly where commercial club releases sit. Streaming will turn it down 5 dB, which is fine — that file was made for a system, not for headphones. |
| -23 LUFS | Broadcast, to EBU R128. | Far quieter than any music context. Only relevant if the audio is going to a radio station or a broadcaster. |
The true-peak ceiling is a separate constraint and it always holds. -1 dBTP is the safe default and it is what most delivery specs ask for; -0.3 is fine for a lossless file that will never be re-encoded; -2 is worth it if you know the track is going to end up as a low-bitrate stream, because lossy encoding pushes peaks up.
Frequently asked
Common questions
Can I normalise a whole folder at once?+
Yes, and it is the reason to use this rather than doing it one track at a time. Drop as many files as you like and one set of settings is applied to all of them, in a queue you can watch — every file gets its own report, so you can see which ones the true-peak ceiling stopped short. The results come back as a single archive rather than as thirty separate downloads. That is the actual job: levelling a night's worth of tracks so you are not riding the gain between them.
Does my file get uploaded?+
No. The file is decoded, measured, gain-changed and written back out as a new file entirely inside your browser tab. Open the network panel and watch it stay empty — there is no upload endpoint behind the page, which also means an unreleased promo is safe to drop on it.
What is the difference between LUFS and dB?+
A decibel measurement of a file usually means its peak — the level of its loudest instant. LUFS is a measurement of how loud the whole thing sounds, weighted for the frequencies the ear is most sensitive to and gated so that quiet passages do not drag the number down. One is a property of a single sample and the other is a property of the music, which is why two files with identical peaks can be far apart in LUFS and why the platforms all use the second one.
Will this make my track louder?+
It will make it the loudness you ask for, if the true-peak ceiling allows. It will not make it louder than that, because getting past the ceiling requires limiting and this tool does not limit — a limiter changes the shape of the music, and doing that silently to somebody's file is not acceptable behaviour for a converter. If the ceiling stops you short, the report says by how much, and then it is a mastering decision.
Does normalising damage the audio?+
A gain change is arithmetic and is essentially lossless — every sample is multiplied by the same number and the relationship between all of them is untouched. The only real cost is at the very end, converting back to a fixed-bit-depth file, and that is handled with dither on 16-bit output. Choose 24-bit or 32-bit float and there is no cost worth naming.
Why is my file quieter than I asked for?+
Because the true-peak ceiling bound first. That happens on tracks with a wide gap between their average level and their peaks — anything with a lot of dynamic range, or a master that was not limited hard. Reaching the target would have pushed the reconstructed waveform over the ceiling, so it stopped, and the report tells you exactly how many units short it was.
Can I normalise a whole set the same way?+
Yes, and it is the right way to prepare one. Run every file to the same loudness target and the gain riding disappears — you are no longer compensating for one track being mastered four decibels hotter than the next. If you want to see what a platform will do to a whole recorded set, the LUFS meter reads the loudness curve over its full length.
Why does it only write WAV and AIFF?+
Because writing an MP3 back out would be a second generation of lossy compression on audio that has already been through one, and that is a real quality loss caused by a tool that was supposed to be fixing your levels. WAV and AIFF are both lossless, both read by every CDJ and by Rekordbox, and AIFF is the one Rekordbox analyses fastest. If you need a smaller file, make it from the lossless output at the end of your chain rather than in the middle of it.
Which bit depth should I pick?+
24-bit for anything you are keeping — it is the working standard, it costs 50% more disk than 16-bit and nothing else. 16-bit if the file is going straight onto a USB stick and space is tight; it is dithered on the way down, so the fade tails stay clean. 32-bit float only if the file is going back into a DAW, where the extra headroom means a later gain change can never clip.
How big a file can it handle?+
Up to 800 MB, which is about 75 minutes of 24-bit stereo audio, so a whole recorded set works. The loudness measurement is a full pass over every sample at four times the sample rate for the peak, so a long file takes a few seconds rather than being instant.
Explore
LUFS meter · Audio quality checker · Audio fade · Silence remover · Waveform generator · LUFS · True peak · Headroom · Limiter · All tools
Build a set from tracks that fit
DJ terms on this page: LUFS · True peak · Headroom · Limiter · Gain normalisation · WAV and AIFF · Sample rate · Waveform · Full glossary