Normalize Audio (LUFS)

Use loudness normalization when you’re tired of the “one clip is quiet, the next is loud” pattern. Instead of matching peak levels, it targets perceived volume measured in LUFS (Loudness Units relative to Full Scale), which correlates much better with how humans actually experience loudness. The result is a file that plays at a predictable, comfortable volume next to other modern audio — and that stands a chance of hitting the loudness target that podcast platforms, streaming services, and broadcasters expect.

Why LUFS instead of peak normalization: a peak normalizer looks at the single loudest sample and scales everything linearly, which means a clip with one loud spike ends up quiet overall, while a densely compressed clip ends up very loud. LUFS looks at the whole file’s perceived loudness, so a quiet podcast and a dense music track can both be brought to, say, −16 LUFS and actually feel similar. For any content that will be consumed back‑to‑back with other audio, LUFS is the right target.

What normalization changes and what it does not: it changes level, not content. It will not remove background noise, de‑echo a room, fix clipping in the original recording, or repair a bad mic position. If the source is noisy, normalization will raise the noise floor alongside the signal — so bringing up a hissy recording to a louder LUFS target also makes the hiss more obvious. Treat normalization as the final balancing step, not a repair pass.

The tool uses a fast one‑pass EBU R128‑style normalization and exports an MP3. One‑pass means it measures and adjusts in a single sweep through the file, which is quick and good enough for almost all speech and podcast workflows. Two‑pass true‑peak limiting in a DAW is more surgical when you need broadcast certification, but for day‑to‑day web publishing the single‑pass approach lands the file in the right loudness ballpark without the extra time cost.

Sensible target levels depend on where the file is going. Spoken‑word podcasts commonly target −16 LUFS (Apple Podcasts’ loose recommendation) or −18 LUFS (looser, more dynamic). Music for streaming platforms tends to sit around −14 LUFS (Spotify). Broadcast content in Europe targets −23 LUFS (EBU R128). For internal sharing or messaging clips where you just want something louder than the original, somewhere between −16 and −14 LUFS is a safe middle ground.

Clipping risk is the main thing to watch. If your source already has peaks near 0 dBFS, pushing the perceived loudness up can force peaks through the ceiling and introduce audible distortion. Two strategies help: pick a lower (quieter) LUFS target so there is more headroom, or run the output through a limiter/clipper in a DAW if you need both high loudness and clean peaks. When in doubt, −16 LUFS is a forgiving target for mixed content.

Typical use cases: balancing multiple podcast episodes so listeners don’t have to adjust volume between them, evening out a conversation where one microphone is much hotter than the other, preparing narration to sit cleanly alongside background music, making field recordings comparable to studio material, and bringing phone voice memos up to something more presentable before sharing. Anywhere different sources are being combined or compared, normalization is the step that makes them feel like a coherent whole.

Workflow order matters. The best practice is: (1) trim and remove unwanted sections, (2) merge your clips in the intended order if you have more than one, (3) normalize loudness on the final file, and only on the final file. Normalizing each clip separately before merging is a common mistake — it produces small loudness differences at every seam because the algorithm reacts to the local content of each clip. Normalize the combined result instead so the whole track is consistent.

What happens when your audio is already loud enough: the tool will reduce gain instead of increasing it. LUFS normalization is bidirectional — if the source measures −10 LUFS and the target is −16 LUFS, the file will be made quieter. This is expected and desirable: platforms usually turn “too loud” submissions down anyway, often with side effects, so delivering at the right level is nearly always better than delivering hot and hoping for the best.

Format and quality notes: output is an MP3 at a sensible bitrate. If you need to normalize losslessly — for mastering workflows, archival purposes, or when the output will be further edited — do the loudness pass in a DAW with a true‑peak limiter, export to WAV or FLAC, and only convert to MP3 at the very end of the chain. For web publishing and podcast delivery, the MP3 produced here is appropriate and doesn’t require that extra ceremony.

Common pitfalls to avoid: do not normalize the same file multiple times in a row expecting it to “get better.” The first pass is already doing the work; additional passes introduce tiny additional lossy re‑encodes without improving loudness match. Do not normalize before heavy editing (cuts, EQ, compression), because subsequent processing changes the loudness you carefully matched. And do not assume normalization will fix a fundamentally uneven recording — structural problems like mismatched mic levels are best addressed with per‑channel gain staging, not with a final LUFS pass.