Audio Editing
LUFS vs dBFS: Audio Loudness and Normalization Explained
LUFS describes loudness measured with a defined weighting and gating procedure. dBFS describes a digital level relative to full scale; when a tool reports sample peak in dBFS, it is reporting the highest sampled amplitude, not how loud the whole programme feels.
By Tubelexity / Updated
Measure loudness and peaks locallyLoudness and peaks are different measurements
LUFS describes loudness measured with a defined weighting and gating procedure. dBFS describes a digital level relative to full scale; when a tool reports sample peak in dBFS, it is reporting the highest sampled amplitude, not how loud the whole programme feels.
A very brief transient and a sustained sound can reach the same sample peak while contributing very different energy over time. This is why turning two files up until their peaks match does not necessarily make a playlist or spoken programme feel consistent.
Integrated loudness, true peak and loudness range
Integrated loudness summarizes a programme over its measured duration using the standard's weighting and gating rules. It is not simply a mean of every silent and non-silent sample. Changing the included introduction, pauses or ending can change the result.
True peak estimates reconstructed waveform peaks between samples and is commonly expressed in dBTP. A decoded lossy export may have higher peaks than its input. Leave headroom and measure the final encoded file, not only the working WAV.
Loudness range summarizes variation in loudness, not the difference between the largest and smallest individual sample. A short steady tone is not a meaningful test of a programme's long-term dynamics. Very quiet or short material may produce unavailable or unstable measurements.
Peak normalization, loudness normalization and compression
Peak normalization applies gain to reach a chosen peak. If the highest sample is -8 dBFS and the target is -1 dBFS, the calculated gain is +7 dB. It raises quiet background noise along with the desired sound; it does not improve recording quality.
Loudness normalization targets an integrated value while respecting a peak constraint. A file with large transients may not reach the desired loudness using a single safe gain change. Limiting or dynamic processing changes the signal, so compare the result instead of assuming the target was met transparently.
Dynamic-range compression applies level-dependent gain using threshold, ratio, attack and release. It can make uneven speech easier to follow, but strong settings can pump, flatten musical attacks or reveal room noise. It is neither file-size compression nor automatic mastering.
| Problem | Start with | Check afterward |
|---|---|---|
| Low peak with otherwise useful balance | Peak normalization | Noise and final peaks |
| Inconsistent programme loudness | Loudness normalization | Integrated loudness and true peak |
| Large level swings within speech | Gentle dynamic compression | Breaths, pumping and intelligibility |
Music, speech and YouTube reference levels
Speech editing often prioritizes intelligibility and consistent phrases. Music may depend on deliberate contrasts between quiet and loud sections. Apply the same measurement vocabulary without assuming that the same compression preset suits both.
EBU R 128 defines a broadcast loudness recommendation; it is not a YouTube upload rule. The commonly discussed -14 LUFS online reference is not an exact YouTube upload requirement established by the sources used here. YouTube's upload-encoding guidance should not be turned into an invented loudness rejection threshold.
Avoid chasing a number by driving a limiter harder. Check how the finished programme sounds on ordinary playback and whether quiet speech remains understandable. EQ can clarify tone; noise reduction may help a noisy source, but neither should be confused with a loudness measurement.
Measure the final delivery, not just the working file
Run the LUFS Meter on the complete source, choose the needed operation, and measure the export again. Compare with the same channel interpretation and programme boundaries. A mono test tone, stereo music mix and dialogue track are not interchangeable calibration material.
Our controlled same-peak experiment below uses synthetic tones to isolate the measurement distinction. It is not a listening study, a platform target test or a benchmark of music quality. Decoders, measurement implementations and programme gating can produce small differences.
Measured example: equal peaks, different integrated loudness
We generated two six-second mono 48 kHz PCM recordings containing a 1 kHz sine. The first uses amplitude 0.25 throughout. The second uses that amplitude for 0.1 second, then 0.025 for the remainder. Both therefore contain the same peak amplitude, but very different sustained energy.
The actual Tubelexity loudness runner measured the following input values with FFmpeg.wasm core 0.12.10. This isolates the distinction between peak and integrated measurement. It does not predict how pleasant music sounds. The very short synthetic material is not suitable for drawing programme loudness-range conclusions.
| Signal | Sample peak dBFS | True peak dBTP | Integrated LUFS |
|---|---|---|---|
| Steady 1 kHz tone | -12.040935 | -12.04 | -15.05 |
| 0.1 s peak, then quieter tone | -12.040935 | -12.04 | -33.95 |
Recommended workflow
- Measure the complete source and note integrated loudness, sample peak and true peak.
- Choose gain, compression or a supported loudness-normalization workflow based on the problem.
- Listen for noise, clipped attacks and pumping.
- Measure the actual delivered file again, including after lossy encoding.