Why can the true peak be higher than the sample peak?

Because samples are instantaneous measurements of a continuous waveform. If the real peak falls between two samples, neither one records it, so a converter that reconstructs the analog signal carefully will output more than the samples suggest, up to about three decibels at the top of the band.

How accurate is the estimate?

It is a four times oversampled band-limited reconstruction, which recovers the classic half-sample crest case cleanly but is not an exact analog simulation. That is why the reading is labelled as estimated and why a delivery ceiling is best left a little below the limit rather than right at it.

Decibels true peak, a scale that is referenced to full scale just like dBFS but measured against the reconstructed waveform rather than the samples. The two scales therefore read the same units but the dBTP figure is always the higher of the pair for a given file.

Why does a loudness target come with a peak target?

Because lossy encoding and conversion can both add inter-sample overshoot, and a master sitting exactly at full scale can distort after a platform encodes it. Specifying a ceiling such as minus one dBTP leaves room for that overshoot, which is why the two specifications travel together.

Does the meter tell me where the peak occurs?

No, only its level. If the position matters, the waveform view shows where the loudest transients sit in the file, and cutting the section out and measuring it separately gives a peak figure for just that passage.

True Peak Meter

You Might Also Need

Peak Limiter

Prevent audio peaks and clipping

Loudness Meter (LUFS)

Measure integrated loudness in LUFS

Loudness Normalizer

Normalize to LUFS loudness standards

Loudness Histogram

Visualize loudness distribution

True Peak Meter compares the peak level of the digital samples with an estimate of the peak that will exist after conversion to analog. The gap between the two figures is where inter-sample clipping hides.

What is True Peak Meter?

Two numbers are reported for the worst channel: sample peak in dBFS, which is what the file's samples literally contain, and an estimated true peak in dBTP, which accounts for the signal that continues between them.

What the panel shows and how the estimate is made:

Sample peak in dBFS and estimated true peak in dBTP, side by side for direct comparison.
Headroom to zero dBTP and headroom to the minus one dBTP delivery target, both reported in decibels.
A verdict line that says whether the file is at or above full scale, below it but above the streaming target, or compliant with the common minus one dBTP target.
The number of channels analysed, noting that the worst channel is the one shown.

The estimate comes from oversampling each channel by a factor of four using a band-limited sinc interpolation with a sixteen-tap Hann kernel, then taking the largest value found. That is a deliberately modest reconstruction, and its behaviour is worth knowing: material close to Nyquist can reconstruct as much as about three decibels hotter than its samples suggest, and the classic test case is a crest that falls exactly halfway between two samples, where the samples read minus three point zero one decibels and the reconstruction recovers close to zero. Linear interpolation between neighbouring samples was tested and rejected because it cannot represent that reconstruction at all, which is why the sinc kernel is used. The result is an estimate rather than a guarantee: a higher oversampling factor or a longer kernel would refine it further, and the tool says so by labelling the reading as estimated. The comparison between the two figures is the useful part: if they are close, the file has little hidden overshoot, and if the estimate runs well above the sample peak, the material is doing something near the top of the band that a converter will reproduce at a level the samples never show. Delivery specifications are written in terms of the estimated figure for exactly that reason, since the ceiling has to hold after encoding and conversion rather than only in the file as stored, whatever sample rate and bit depth it happens to use.

Where the measurement has limits:

The verdict is based on the worst channel, so a per-channel breakdown is not shown.
The reconstruction is four times oversampled, so a true peak finer than that is not resolved exactly.
Only the peak is reported, so loudness has to be measured separately.
Nothing is limited or corrected: an over-ceiling file needs the limiter or a gain change.
Inter-sample estimation cannot predict what a specific digital-to-analog converter will do, only how much hotter the reconstructed signal may be.
No log of peak positions in time is produced.

Use it as the last check before delivery, because lossy encoding and digital-to-analog conversion can both push a signal above the ceiling its samples appear to respect. Set the output ceiling with a limiter and confirm with this meter rather than trusting the sample peak alone.

How to use True Peak Meter

Drop an audio file onto the panel; every channel is analysed and the worst one is reported.
Compare sample peak against the estimated true peak to see how much is hidden between the samples.
Read the headroom figures against zero dBTP and against the minus one dBTP delivery target.
If the verdict warns about full scale, reduce gain or apply a limiter and measure again.

When to use True Peak Meter vs related tools

True Peak Meter is the delivery ceiling check, so it belongs alongside loudness measurement rather than instead of it. Loudness Meter (LUFS) supplies the loudness side of the same delivery question, and Peak Limiter or Brickwall Limiter is where a ceiling is actually imposed. To see whether the peaks are unusual for the material, Crest Factor Meter describes their relationship to the average level.

Privacy & Security

This tool runs entirely in your browser — no data ever leaves your device. There is no server round-trip, no upload, no logging, and no account required. Your input is processed locally using client-side JavaScript and is never stored, transmitted, or accessible to anyone else. When you close the tab, everything disappears.