True Peak Explained

A track can show 0 dBFS peak in your DAW's meter, look perfectly safe, and still clip audibly once it's converted to a lossy format or played back through a real digital-to-analog converter. The reason is the gap between what a sample-peak meter measures and what actually happens to the waveform between samples.

Sample peak only looks at the dots

Digital audio is a sequence of discrete samples — at 44.1kHz, 44,100 numbers per second per channel. A standard peak meter reports the single largest of those numbers. That's sample peak, and it's an honest measurement of the data — but it isn't the whole waveform.

What happens between the dots

When that digital signal gets converted back to an analog voltage (or resampled, or transcoded to a lossy codec), a reconstruction filter draws a smooth continuous curve through the sample points. That curve can overshoot the highest individual sample value — particularly around closely-spaced samples that are both near full scale but not perfectly aligned. The result: the actual reconstructed waveform peaks higher than any single sample ever showed, sometimes enough to clip in the analog domain, or push a lossy encoder's internal peak over its own ceiling, even though every sample in the file was technically "under 0 dBFS."

True peak is the estimate of that reconstructed, inter-sample peak — expressed in dBTP (decibels, True Peak) rather than plain dBFS, to make clear it's a different, higher-fidelity measurement than sample peak.

How it's actually estimated

You can't perfectly reconstruct the analog waveform from software, but you can get a close, standard-compliant estimate: oversample the signal — insert and interpolate extra samples between the real ones — then measure the peak of that oversampled version. This site's analyzer oversamples by 4x specifically for this measurement, which is the same general approach ITU-R BS.1770 specifies for true-peak estimation. It's explicitly an estimate, not a mathematically exact reconstruction — a higher oversampling factor gets closer to the true analog peak at the cost of more computation, and 4x is a standard, practical balance point.

What the number means in practice

• True peak > 0 dBTP means the estimated reconstructed waveform exceeds full scale — real intersample-clipping risk on playback or transcode, even with every sample in the file under 0 dBFS.
• -1 dBTP is a commonly recommended safety ceiling for delivery, giving margin below the hard 0 dBTP risk point specifically to absorb the small peak increases that lossy transcoding (MP3, AAC, Opus) can introduce on top of whatever the file already measures.
• Streaming platforms vary in exactly what they recommend or enforce, and that guidance shifts over time — treat any single number you read as current guidance, not a fixed law, and build in margin rather than mastering right up against whatever the number happens to be today.

Check your own track

Analyze My Track reports true peak in dBTP alongside integrated LUFS, short-term/momentary loudness, and dynamic range — on your actual upload, so you can see exactly how much true-peak margin your master actually has, not an assumption based on the sample-peak reading your DAW showed you.