HERTZKIT / AUDIO PRACTICE
LUFS, dBFS and true peak: what you measure
Distinguish loudness, sample peak and true peak to interpret measurements and prepare a delivery.
On this page7 sections
In the LUFS meter, choose a file or Try a piano example, then choose Measure loudness. It measures the whole track; playback position and volume do not change the readings.
Hear two versions matched for LUFS.
Review the integrated loudness, peaks and chart. Select a moment on the chart and use Listen from here to investigate it. On a phone, View results and Back to audio connect the two panels.
Compare a reference is optional. Enable Simulate gain and include it in the report to compare uniform gain against a chosen peak ceiling. No gain is applied to audio. Export report includes the measurements and chart; the simulation appears only when enabled. Print or save as PDF, or download TXT and CSV.
Canceling a measurement keeps the audio ready to retry. References are comparisons, not universal mastering targets. Processing or encoding can change the peak: measure the final exported file.
Peak and loudness answer different questions
Sample peak indicates the largest digital sample value, expressed in dBFS. LUFS uses weighting and time analysis to estimate loudness; integrated loudness summarizes a program with its gating system. Two files with the same peak can have very different loudness. Do not infer how loud a song will sound by looking only at its peak.
What does true peak add?
True peak estimates the reconstructed signal peak between samples and is expressed in dBTP. It can exceed sample peak. Meters and oversampling methods can differ, especially near the limit. HertzKit provides an estimate for working and comparing; it is not a certified meter for verifying a regulated delivery. For a strict requirement, check with a validated meter and listen to the final file.
There is no single target for everything
EBU R 128 recommends −23 LUFS for its program-normalization context; it is not a universal requirement for online music. Platforms and recipients may use other conditions. Check their current specifications. Measure the entire file when assessing integrated loudness, preserve the dynamics the music needs, and do not confuse reaching a number with improving a mix. To compare two versions, match loudness and listen to the same passage.
Two versions: a numerical example
Imagine two exports of the same passage: A measures −18 integrated LUFS and B measures −14 integrated LUFS. B has a measured loudness advantage of 4 LU. For an initial comparison, turning B down by about 4 dB approximately removes that advantage. You are not choosing −18 as a release target: you are preparing a fairer listen. The comparator can match listening levels without changing your original files.
Version A
- Integrated loudness
- −18 LUFS
- Sample peak
- −1 dBFS
Version B
- Integrated loudness
- −14 LUFS
- Sample peak
- −3 dBFS
Here B has 4 LU more integrated loudness despite a lower sample peak. Matching peaks does not necessarily match loudness. These figures illustrate a comparison; they are not measurements of your files.
What to check before delivery
- Measure the entire final file, including its ending; do not use only the chorus to describe the whole song.
- Read integrated LUFS, sample peak and true peak separately. After changing gain or limiting, measure the export again.
- Check the recipient’s specification. If it requests a ceiling in dBTP, a sample peak value in dBFS is not a substitute.
- Listen to the actual file you will send: beginning, transitions and ending. Correct measurements do not identify a cut-off word or an out-of-time edit for you.
Put the idea into practice
Check LUFS with two audio files
Download the piano study (10 s, stereo) and the Sessions reference (4.5 s, mono). Open each one in the Audio Analyzer and note the integrated loudness of the full file.
With FFmpeg installed, repeat the measurement. Replace audio.wav with each filename and read I in the final summary.
ffmpeg -hide_banner -nostats -i audio.wav -af ebur128 -f null -- Piano: −21.7 LUFS.
- Sessions reference: −19.6 LUFS.
HertzKit’s calculation and FFmpeg 7.1.5 agreed to one decimal place on these two files. This check compares integrated loudness on mono and stereo examples. The readings describe these recordings and are independent of any delivery target.