What a LUFS meter tells you that a peak meter can't
Peak meters show how close the waveform gets to 0 dBFS, but two files with identical peaks can sound wildly different in volume. LUFS (Loudness Units relative to Full Scale, also written LKFS) measures how loud audio sounds to people. The signal is first passed through a K-weighting filter that mimics how our hearing favors the upper mids and ignores deep bass, then its energy is averaged over time.
Integrated loudness is that average across the whole file. Following BS.1770-4, near-silence below −70 LUFS is ignored, and so is anything more than 10 LU quieter than the rest, so pauses between sentences don't drag a podcast's number down. One LU is the same step as one decibel, which makes the gap to a target easy to read: a file at −17 LUFS needs 3 dB more to reach −14.
Reading the numbers
- Integrated (LUFS), the single number platforms normalize to. Use it to compare against a target.
- True peak (dBTP), the highest level the waveform reaches between samples, estimated by oversampling. MP3 and AAC encoding can push those peaks higher, so most specs ask for −1 dBTP or lower.
- Loudness range (LU), how much the loudness swings, ignoring the extremes (EBU Tech 3342). Speech usually sits around 3 to 8 LU; a dynamic film mix can exceed 15.
- Max short-term and max momentary, the loudest 3-second and 0.4-second stretches. Handy for spotting a single blast in an otherwise calm recording.
The loudness over time graph plots short-term loudness every half second, with your integrated level and the selected target drawn across it. Sections that sit far above the line are the ones that will feel jarring.
Loudness targets for streaming, podcasts and broadcast
- Spotify normalizes playback to −14 LUFS and recommends true peaks below −1 dBTP. YouTube doesn't publish a target, but its normalization measures out at about the same level. On both, louder uploads are turned down, so mastering hotter gains nothing.
- Apple Podcasts asks for −16 LKFS (±1 dB) with true peaks no higher than −1 dB, a sensible target for spoken word anywhere.
- Mono podcasts: some producers aim about 3 dB lower (around −19 LUFS), because a mono file played through two speakers sounds louder than its reading suggests.
- Broadcast: EBU R128 in Europe asks for −23 LUFS with peaks under −1 dBTP; ATSC A/85 in the US asks for −24 LKFS with peaks under −2 dBTP.
“On target” in the list means within 1 LU. Delivering a broadcast master? Check your broadcaster's own tolerance: R128 itself allows ±0.5 LU for most programmes.
How normalizing works
Normalizing applies one fixed gain change to the whole file, so the mix and its dynamics stay exactly as they were, only the overall level moves. When turning up would push peaks past your ceiling, a transparent look-ahead limiter catches just those peaks. The panel tells you beforehand how much limiting to expect; if it's more than a few decibels, a slightly lower target will keep more punch.
After saving, ClipLime decodes the new file and measures it again, so the before-and-after numbers describe the file you actually download, including any peak overshoot added by MP3 or M4A encoding. For the strictest delivery specs, save as WAV, which can't overshoot.
Need to even out loud and quiet passages first, or cut a section? Do that in the audio editor, then come back to check the final number. To change formats without touching the level, use the audio converter.