Normalize Audio (Peak and LUFS)

Normalization in plain terms

To normalize audio means to find out how strong a file is and then apply a single, fixed gain so that it reaches a chosen target. Nothing else is changed. There is no compression, no equalizing and no limiting: every part of the recording moves up or down by the same number of decibels, so the balance between loud and quiet moments stays exactly as it was recorded.

That single-gain idea is what separates normalizing from other level tools. A compressor reshapes dynamics. A limiter squashes peaks. Normalization is a measurement followed by one multiplication, which is why it is safe to use on speech, music or a field recording without altering the character of the sound.

The interesting question is what you measure. WaveTinker offers two answers. Peak mode looks at the single highest sample in the file and sets that to a level you choose. Loudness mode measures how strong the whole file sounds over time, in LUFS, and sets that to a target. They solve different problems, and picking the right one is most of the work.

Peak or loudness: which mode to use

Use Peak (dBFS) when the goal is to make the best use of the available range without risk of overshoot. A quiet recording that peaks at -18 dBFS can be raised by 17 dB so that its highest point lands at -1 dBFS. This is the traditional idea of normalizing, and it suits a single clip, a sound effect, a sample for a music project, or the last step before sending a file to someone who will mix it.

The weakness of peak mode is that the highest sample says little about how strong a file feels. A single sharp click can set the peak of an otherwise quiet interview, and two files normalized to the same peak can sound very different in strength. If you line up a podcast and a pop song at -1 dBFS peak, the song will feel far stronger, because its energy is packed more tightly.

Use Loudness (LUFS) when files need to feel similar to a listener: a podcast season, a playlist of recordings from different sources, or audio headed for a platform that matches level to a target. LUFS follows how human hearing weights frequencies and ignores silence, so it is a far better predictor of perceived strength than the peak is.

What LUFS means and how the tool measures it

LUFS stands for loudness units relative to full scale. One unit is one decibel, so a change of 3 LU is a change of 3 dB. The measurement follows ITU-R BS.1770: the audio passes through a K-weighting filter that mimics how we hear, is cut into overlapping blocks of 400 milliseconds, and quiet blocks below a threshold are gated out so that pauses do not drag the average down. The result is the integrated loudness of the entire file, a single number.

Alongside it, the tool shows the sample peak. This is the highest sample value in the file, which is not quite the same as the true peak. A true-peak meter estimates the level that can appear between samples after conversion to analog, and it can read a little higher than the sample peak. WaveTinker reports sample peak only, so leave a margin if a platform demands a strict true-peak ceiling.

Loudness figures for short material should be read with care. A three-second effect has too little content for a stable average. Treat the number as meaningful for anything that lasts at least several seconds, and as a rough guide below that.

Targets: common loudness values

The table in the Numbers tab lists common targets. Spotify's default setting is -14 LUFS, and YouTube sits at about the same figure according to measurements by engineers. Apple Music's Sound Check lands near -16, which is also a frequent recommendation for podcasts. European broadcasters follow EBU R 128 at -23 LUFS, and US broadcast follows ATSC A/85 at -24 LKFS, a unit that is numerically the same as LUFS.

Streaming platforms mostly turn tracks down when they are above their target, and in some cases raise them when below. That is why chasing a very hot master rarely pays. The service will reduce it, and the result may sound flatter than a less processed version. Because platforms change their settings from time to time, treat the table as a guide and check current documentation before preparing material for a specific destination.

The presets mirror this list: Streaming -14 LUFS, Podcast -16 LUFS and Broadcast -23 LUFS in loudness mode, plus Peak -1 dBFS and Peak 0 dBFS in peak mode. The Target knob turns to any value from -30 to 0 in half-decibel steps if your requirement is different. Inside the studio on the home page the Loudness module follows the same logic (LUFS only, from -30 to -6), and its Podcast polish recipe starts from -16.

When the gain is held back

Hitting a loudness target by gain alone can push peaks past full scale. Suppose a dynamic piece measures -22 LUFS with peaks at -4 dBFS, and you ask for -14 LUFS. That needs +8 dB, and the peaks would land at +4 dBFS, which cannot exist in a digital file.

In loudness mode the tool handles this by holding the gain back so that peaks stay under -1 dBFS, and it says so on the page. The result will then sit below your target, and the readout shows by how much. That is the honest outcome of a single gain: it cannot invent headroom that the recording does not have.

You have a few options when this happens. Accept the lower figure, which is often fine because a platform will not punish a file for being a little under target. Use the volume booster, which has a limiter, accepting that this changes the dynamics. Or go back to the source and look for a version with less extreme peaks. What the normalizer will not do is quietly clip your file.

Reading the before and after figures

After you choose a target, the page shows measurements before and after processing. Before is the file as it came in: its integrated loudness and its sample peak. After is the same pair of numbers for the result. Comparing them tells you what happened without any guesswork: the difference in LUFS is the gain that was applied, and the new peak shows how much room remains.

If the after-loudness equals your target and the peak is comfortably under -1 dBFS, nothing more is needed. If the loudness falls short and a note says the gain was held back, the peak is what is limiting you. If you choose Peak mode, the after-peak will equal your chosen ceiling, and the loudness figure simply reports where that ended up.

A useful check is to flip between Result and Original. They switch at the same spot in the track, so you hear the effect of the gain alone. The space bar plays and pauses. If the Result seems much stronger, remember that the human ear tends to prefer the stronger version in a quick comparison, which is a reason to trust the numbers and not only the impression.

Practical recipes

A podcast episode recorded by two people on different microphones: normalize each speaker's track to -16 LUFS before mixing, so neither dominates, then check the final mix against the same target.

A set of voice memos for a project: normalize each to -16 LUFS in loudness mode so that playing them in sequence does not require reaching for a control.

A playlist of tracks gathered from several albums that you own: normalize each to the same LUFS target so the sequence plays evenly in a player without a matching setting.

A single sound effect for a game or video: use Peak with -1 dBFS to place it just under full scale and leave a margin for later processing.

A file that needs to meet a broadcast spec: choose the -23 LUFS preset, then check the peak line, since broadcasters often have a peak ceiling as well.

Saving, formats and privacy

Open MP3, WAV, M4A and AAC, OGG and Opus or FLAC files of up to 250 MB. The result can be saved as MP3 at 128, 192, 256 or 320 kbps or as WAV in 16 or 24 bits, with a size estimate shown first. Because lossy encoding can nudge peaks slightly, leaving the -1 dBFS margin is wise when the final file is an MP3. The tool works on files already on your device and does not retrieve audio from any website or service. Videos are accepted as sources too: the sound of most MP4, MOV and WebM files is measured and leveled, and the output is audio.

Measurement, gain and encoding are all carried out locally in the browser tab. The audio is not uploaded, and no copy exists on a server. Settings can be shared with the Copy settings link button, which places only the mode and the target in the page address.