Match loudness levels and remove volume jumps
Normalize MP3 and WAV files to consistent loudness with local processing.
Privacy First: All processing runs locally in your browser. Your files never leave your device.
Drag and drop your file here
or click to browse
Supported: MP3, WAV | Max file size: 10MB | Max duration: 5 minutes
Files to merge (0)
Use arrows to reorder, × to remove files.
Processing happens locally for privacy.
No uploads. Your files stay local.
Normalize in seconds with minimal steps.
No registration required.
MP3, WAV.
Playlists recorded from different sources jump in loudness — one track whispers, the next blasts. Normalizing brings every file to a consistent target level so you never touch the volume knob mid-playlist. Run each file through the normalizer with the same target and they'll match.
Podcast platforms expect consistent loudness (typically around −16 LUFS for stereo). Normalize your episode before uploading and listeners won't be blown out by ads or straining to hear quiet passages. Pair it with the silence remover for gaps and the audio joiner to stitch segments — or let the AI jingle maker handle mixing and mastering of your intro automatically.
Every platform normalises playback to its own target, so the only sensible approach is to deliver at the target rather than as loud as possible. Mastering hotter than the target does not make you louder — it gets turned down, and you keep the squashed dynamics you paid for it with.
| Destination | Target | True peak | Why |
|---|---|---|---|
| Spotify, YouTube, most streaming | −14 LUFS | −1 dBTP | Both normalise playback to roughly this level |
| Podcasts, Apple Music | −16 LUFS | −1 dBTP | Speech needs headroom for music beds and ads |
| Broadcast radio and TV (EBU R 128) | −23 LUFS | −1 dBTP | The level European broadcasters deliver to |
| A station's own playout / imaging | Match the library | −1 dBTP | An ID 3 dB hotter than the music is the usual complaint |
Two rules that save most of the trouble. Keep the true-peak limit at −1 dBTP whatever the target, because a lossy encoder can overshoot a measurement taken before encoding. And normalise last: cut the silence, trim the ends and join the segments first, then set the level on the finished file — a joined file made of differently levelled parts cannot be fixed by one gain change.
If your file measures −14 LUFS and still sounds smaller than a commercial release at −14 LUFS, the difference is dynamics, not level. A commercial master has been compressed, so its average sits close to its peaks and it feels dense all the way through. A dynamic recording with the same average has much taller peaks and much quieter valleys, so it reads quieter even though a meter says otherwise.
The fix is compression before normalisation, not a hotter target — pushing the target up just asks the limiter to flatten your loudest moments. Measure both files in our audio analyzer and compare the crest factor (peak minus RMS): under about 8 dB is a squashed modern master, 10–14 dB is a healthy mix, and over 20 dB is a recording that will disappear in a car. That number explains the loudness difference far better than the level does.
One more thing worth knowing: raising a quiet recording raises its noise floor with it. If a file needs more than about 12 dB of gain, the hiss comes up audibly, and normalising is the wrong tool — re-record it if you can, or accept it.
Frederick — founder of AI Jingle Maker and a professional voiceover artist for the past 25 years, working from London. He builds the tools on this site, writes these pages, and still records and mixes the handcrafted jingle packages himself in FL Studio using his own library of music and effects. There is no content team here: if something on this page is wrong, it is his mistake to fix, and you can reach him directly.
Radio World profiled him and these tools in its News Makers section — Using AI to Offer a Big-Market Sound on a Small-Market Budget (25 July 2026).
Yes, it is free with no signup.
Normalizing changes level, not content: it applies gain so the whole file lands on your loudness target, and the true-peak limit stops the loudest moments going over. Nothing is re-balanced and no frequencies are touched. The one thing to know is that raising a very quiet recording also raises its noise floor by the same amount — hiss and room noise come up with the voice — so normalising is a finishing step, not a rescue for a badly recorded file.
No, processing is local.
MP3 and WAV are supported.
Match the destination. −14 LUFS for Spotify, YouTube and most streaming; −16 LUFS for podcasts and for Apple Music; −23 LUFS if the file is going to a broadcaster working to EBU R 128. Keep the true-peak limit at −1 dBTP in every case: lossy encoders can overshoot by a fraction of a decibel, and that margin is what stops a file that measured clean from clipping on playback.
Peak normalisation finds the single loudest sample and scales the file so that sample sits at, say, −1 dBFS. It guarantees no clipping but says nothing about how loud the file feels — one stray transient can hold a whole quiet recording down. Loudness normalisation works on the average the ear actually responds to, measured in LUFS with the K-weighting and gating defined in ITU-R BS.1770, and that is what every streaming platform uses. This tool works to a LUFS target with a peak limit on top, which is the combination you want.
Because loudness is not only level. A commercial master is compressed, so its average sits close to its peaks and it can hit a target without any single moment sticking out. A dynamic recording normalised to the same target has the same average but much bigger peaks, so it feels quieter. If you need it to compete, the answer is compression and limiting before normalising — not a higher target, which only invites the limiter to flatten the loud parts.
Yes, as long as you use the same target for all of them: that is the point of a loudness target rather than a peak target. Normalise each file to −16 LUFS and they will sit at the same perceived level in a playlist, even if they were recorded in different rooms on different devices. Join them afterwards with the free audio joiner — levelling first, joining second, because fixing a level inside a merged file means editing by hand.
Turn your idea into a finished jingle, radio ID, or promo. Make one yourself in seconds with our AI Jingle Maker — or have a custom one handcrafted for you.