Normalize Audio

What is LUFS, and why not just use peaks?

LUFS — Loudness Units relative to Full Scale — measures perceived loudness over time, weighting frequencies the way human hearing does. That is why it beats peak-based methods: two files can have identical peaks yet feel wildly different in loudness because one is dense and one is sparse. Peak normalization only guarantees the single loudest sample hits a ceiling; LUFS normalization (EBU R128) matches the average loudness you actually experience, which is what makes tracks stop lurching between quiet and loud.

There are three LUFS readings worth distinguishing. Integrated loudness is the single number for the whole file — the one you normalize to a target. Short-term and momentary loudness describe how loud the file is over the last few seconds and the last fraction of a second, useful for spotting passages that spike. Alongside them sits true peak, the highest instantaneous sample level, which a good normalizer keeps under the ceiling so that hitting a loudness target never sneaks clipping back in.

Normalize vs. compress vs. amplify — what is the difference?

These three get muddled constantly, so it is worth pinning down:

• Normalize measures the file’s loudness and raises or lowers it so the average hits a target (here −16 LUFS). It is about level, not about squeezing dynamics the way a compressor does.
• Amplify (adjust volume) applies a gain you type by hand — no measurement, no target. Blunt and manual.
• Compress (dynamic-range compression) reduces the gap between loud and quiet *within* the file, evening out moment-to-moment swings. Note: "compress" here means dynamics, not file size — that other meaning of compress lives on the compress audio page.

In practice the three often work together. You might compress a voice recording to tame the gap between loud and quiet phrases, then normalize the compressed result to −16 LUFS so it sits at the same level as everything else in the feed. Amplify is what you reach for only when you already know the exact decibels you want and have no need for a measured target at all.

Does normalizing upload my audio?

No. Both the loudness measurement and the gain correction run in ffmpeg.wasm inside your browser. Your file is analysed and rewritten locally and never sent anywhere — confirm it in DevTools → Network, where normalizing produces no upload. Unreleased mixes and confidential recordings stay on your machine.

Related audio and video tools

Need a manual, exact dB change instead of a target? Use adjust audio volume. Smoothing the ends is the job of fade in and out. Once loudness is even, compress the file for delivery. And if the audio is currently inside a clip, extract the audio from the video first, then normalize it here.