Every file at the same loudness. Nothing uploads.

Measure any audio file's real loudness in LUFS, then normalize it to streaming, podcast or broadcast level — one file or a whole batch, with the after numbers measured from the actual download. Entirely in your browser.

  • Real LUFS, measured before & after
  • Batch up to 5 files + album mode
  • WAV or MP3 output
  • Free, no account, no uploads

Normalize audio volume to a LUFS target

Drop audio files here, or click to browse

MP3, WAV, FLAC, M4A, AAC, OGG, Opus, AIFF · up to 200 MB each · up to 5 files · up to 2 hours each

Loudness target

Presets carry their own peak ceiling (podcast: -1.5 dBTP, the rest: -1 dBTP). Custom lets you set both.

Batch mode

Per file is right for unrelated clips. Album is right when the files belong together — an album, a multi-part recording, a set of takes.

Output format

WAV keeps every sample; MP3 is lossy but much smaller and plays everywhere. Tags are carried over where present.

Drop an audio file above — its loudness is measured automatically, then press Normalize audio.

No files yet — drop audio above, or try the sample tone.

Privacy you can verify, not just trust

Most loudness tools send your audio to a server, or cap free files per day and charge for the rest. This one can't: the audio engine (ffmpeg.wasm) is fetched once from this same site, and every measurement and every normalization runs in your browser. Your audio never leaves the device — the only third-party requests are the site's own analytics, which count page views — and it uploads nothing, ever. Don't take our word for it:

New cross-origin requests during use: 0
  1. Press F12 to open developer tools, then the Network tab.
  2. Press Try a sample tone and normalize it above. Watch: apart from the site's analytics, nothing new is requested — no uploads, and no audio ever leaves your device.
  3. Extra credit: disconnect your internet after this page loads — measuring and normalizing still work once the engine has loaded.

How to normalize audio volume to a LUFS target

Streaming platforms play every track at roughly the same loudness — about -14 LUFS on Spotify and YouTube, about -16 on Apple Music — and turn louder tracks down without asking. A podcast recorded at -26 LUFS plays quietly next to one recorded at -16, and a master at -8 LUFS gets turned down anyway. Normalizing before you publish means the platform plays your audio the way you made it, instead of deciding for you.

To use it:

  1. Drop your audio files — MP3, WAV, FLAC, M4A, AAC, OGG, Opus or AIFF, up to 5 at a time. Each file is measured automatically: integrated loudness in LUFS, true peak in dBTP and loudness range, shown on its card before anything changes.
  2. Pick a target. Streaming (-14 LUFS) for music and video, Podcast (-16 LUFS) for spoken word, Broadcast (-23 LUFS) for EBU R128 work, or a custom level from -30 to -6 LUFS with your own true-peak ceiling.
  3. Pick a batch mode. Per file normalizes every file to the target on its own. Album applies one gain to the whole batch, so an album or a set of takes keeps its relative balance — the loudest track lands on target and the rest follow.
  4. Press Normalize audio. The engine normalizes each file, then re-measures the finished file — the after numbers on the card are measured from the actual download, not from the normalize pass's own estimate. Download files one at a time, or all of them as a ZIP.

What makes this normalizer different

The pages that top the results for this search are good tools, and most of them do one part of the job well. Pi7's audio normalizer batches files and re-measures them, but only exports the format you put in. AudioUtils exports MP3 only and caps files at 20 MB. GrepCut and MixMasterAI export WAV only. KlipTools takes one file at a time. This page combines the parts the others split up: batch normalization with per-file before-and-after LUFS measured from the actual output, an album mode that keeps a batch's relative balance, both WAV (lossless 16-bit PCM) and MP3 (192 kbps) output, a built-in sample tone so you can try it without a file, and a live network-request counter so you can verify the privacy claim yourself. No file-size cap below 200 MB per file, no account, no daily limit.

When is this the right tool?

Whenever audio from different sources has to play at the same loudness: podcast episodes recorded on different days, music tracks for a playlist or a release, voiceovers cut in different sessions, video clips for one channel, or a batch of samples for a sound library. If the files belong together as a set, use album mode; if they are unrelated, use per-file mode.

Honest limitations

Single-pass accuracy is about ±1 LU

Per-file mode uses ffmpeg's loudnorm in one dynamic pass, which is accurate to roughly ±1 LU. The after number on each card is re-measured from the finished file, so it is the real result — and when the true-peak ceiling caps the gain, the file lands below target instead of clipping, and the card says so.

MP3 output is lossy

MP3 output is libmp3lame at 192 kbps — small and playable everywhere, but lossy. If the file is a master you will edit again, download WAV instead. WAV output is 16-bit PCM; a 24-bit source is dithered down to 16-bit by the encoder.

It is not a compressor or a limiter

Album mode applies one static gain and nothing else. Per-file mode's loudnorm pass can gently limit peaks to hold the ceiling, but it does not compress dynamics, remove noise, or fix clipping that is already in the recording. A file that clips at the source still clips after normalizing — it just clips quieter.

Long files take a while

In-browser processing runs at roughly real-time speed, and each file goes through three passes (measure, normalize, re-measure). A 2-hour file is the limit, and anything over 30 minutes warns you first. For a whole library, a desktop normalizer will be much faster.

Silence measures as nothing

A file that is mostly silence, or a file with a long quiet intro, measures by its gated loudness — near-silence is discarded by the measurement, so the number reflects the audible part. A fully silent file has no loudness to measure and is refused with an explanation instead of a made-up number.

Surround and odd formats

Stereo and mono are fully supported. Multichannel files keep their channels in WAV output; MP3 output is stereo. Video files are not accepted — strip the audio to MP3 or WAV first, or use a video converter.

Frequently asked questions

What LUFS target should I pick?

Use -14 LUFS for music and video going to Spotify, YouTube or Tidal, which play at about that level. Use -16 LUFS for podcasts and spoken word — Apple Podcasts asks for about -16 LKFS with true peak no higher than -1 dBFS. Use -23 LUFS for broadcast work under EBU R128. Use -10 LUFS only for clips going somewhere with no normalization at all, such as SoundCloud or a club PA, where masters sit louder. If you are unsure, -14 LUFS is the safe default for music and -16 LUFS for voice.

Does normalizing compress my audio or change how it sounds?

Per-file mode uses ffmpeg's loudnorm filter in single-pass dynamic mode. It measures the whole file's perceived loudness and moves it to the target, holding true peaks under the ceiling you pick. It is not a compressor — it does not squash the loud and quiet parts together — but the dynamic pass can gently limit peaks, so a very loud file may land slightly under target rather than clip. Album mode is different: it applies one static gain to every file, so it changes nothing about the sound except the volume. The after numbers shown for every file are re-measured from the finished download, so you can see exactly what happened.

What audio formats can I normalize, and what do I get back?

You can drop in MP3, WAV, FLAC, M4A, AAC, OGG, Opus and AIFF files — up to 200 MB each, up to 5 files per batch, up to 2 hours per file. You choose the output: WAV (16-bit PCM, lossless) or MP3 (192 kbps, for small files and easy sharing). Tags such as title, artist and album are carried over where present; any cover art is dropped. If a file is actually a different format than its name says — a WAV renamed to .mp3, for example — the tool tells you instead of failing silently.

Does my audio upload anywhere?

No. Your audio is processed entirely in your browser with ffmpeg.wasm — the engine files are fetched once from this same site, and your audio never leaves your device. The only third-party requests are the site's own analytics, which count page views. You can verify it yourself with the network-request counter in the privacy section below the normalizer.

Why did my file come back quieter or louder than the target?

Two honest reasons. First, the true-peak ceiling: if reaching the target would push a peak over the ceiling (default -1 dBTP), the gain is capped and the file lands below target instead of clipping — the file card says so when it happens. Second, single-pass loudnorm is approximate, usually within about 1 LU of the target. The after number shown is re-measured from the finished file, so it is the real result, not a promise. If you need an exact hit, lower the target by the measured error and run it again — do not raise the ceiling.

What is the difference between per-file mode and album mode?

Per-file mode normalizes every file to the target on its own, so a quiet podcast episode and a loud one both land at -16 LUFS. Album mode applies one gain to the whole batch: the loudest file lands on target and every other file sits below it by the same amount it did before, so an album, a multi-part recording or a set of takes keeps its relative balance. Use per-file mode for unrelated clips and album mode when the files belong together.

What is true peak, and why is there a ceiling?

True peak is the highest level the audio reaches between samples, once the waveform is reconstructed the way a player does it — it can be higher than the highest sample value. Streaming platforms ask for true peaks below -1 dBTP because anything at 0 dBFS can clip on playback after encoding. The ceiling (default -1 dBTP, -1.5 for the podcast preset) stops the normalizer from boosting a file into clipping. If a file is already very loud, the ceiling is what keeps the output clean.

Is there a file size or number limit?

Each file can be up to 200 MB and up to 2 hours long, and you can normalize up to 5 files per batch. Files over 30 minutes get a warning first, because in-browser processing runs at roughly real-time speed. There is no account, no daily cap and no watermark — normalize in smaller batches if you have a whole library.