True Peak Explained: Why Platforms Ask for -1 dBTP

Your file peaks at -0.3 dB in the editor. The meter says it never clipped. Then a listener hears a crackle on the loudest laugh. The missing idea is true peak: the level of the sound between the samples your editor shows you. This guide explains what true peak is, why every platform caps it at -1 dBTP, and how to stay under the ceiling.

When you normalize with LoudNorm, the true-peak ceiling is applied with the loudness target — the report shows both numbers, before and after.

Samples Are Not the Whole Wave

Digital audio is a row of snapshots. Between two snapshots, the real wave can rise higher than either of them — like a ball passing between two fence posts can be higher than both posts. A normal peak meter reads the posts. A true-peak meter reconstructs the wave between them, the way a converter or a phone speaker will. That reconstructed peak is the true peak, written in dBTP.

Put numbers on it. At a 44.1 kHz sample rate, your file takes 44,100 snapshots per second. That sounds like a lot, but a sharp transient — a consonant, a snare hit, a laugh's attack — can crest between two snapshots that each read lower. Oversampling meters (the BS.1770 method uses at least 4× oversampling) rebuild the curve and find the real top. Overshoots of 0.5–1 dB over the highest sample are common in hot masters, and larger ones happen: a sample peak of -0.3 dBFS with a true peak of +0.8 dBTP is not rare in audio that was limited hard against 0 dB.

ReadingWhat it tells you
Sample peak -0.3 dBFSThe tallest snapshot in the file
True peak +0.8 dBTPThe tallest point of the real wave — over full scale
What the listener getsA converter asked to output a level it cannot reach: distortion

The file "never clipped" in the editor and clips at the listener's house. Both statements are true; they are just measuring different waves.

Why -1 dBTP Is the Line

Two things happen to your file after upload. It gets re-encoded (MP3, AAC), and encoding nudges peaks upward. Then it gets played through converters that reconstruct the wave. A file whose true peak sits at 0 dBTP can overshoot on both steps and clip — the crackle your listener heard. Leaving 1 dB of headroom (-1 dBTP) absorbs both nudges. That is why Apple, Spotify and the broadcast rules all ask for it, alongside their LUFS targets in the LUFS explainer.

Why encoding raises peaks: MP3 and AAC do not store the waveform itself; they store a recipe for rebuilding something close to it. The rebuilt wave is close, not identical, and "close" around a sharp peak often lands a little higher than the original. The louder and more limited the master, the bigger the drift. A master parked at -1 dBTP has room for that drift. A master at 0 dBTP converts the drift into distortion on your loudest, most important moments — usually the laugh, the chorus hit, the one line the episode is remembered for.

Some engineers use -2 dBTP for extra safety, especially for low-bitrate destinations. It costs almost nothing in loudness — the loudness target is an average, and shaving the rarest peaks by an extra decibel barely moves it. When in doubt, the safer ceiling wins.

How to Stay Under the Ceiling

  1. Measure true peak, not sample peak. Any BS.1770 loudness meter reports it.
  2. Leave the ceiling in the normalization. When you normalize to a loudness target, set the true-peak limit to -1 dBTP at the same time; the two settings travel together.
  3. Expect a trade-off on hot masters. If a file is already loud and peaky, holding -1 dBTP may keep it a little under the loudness target. A good report shows both numbers honestly — see how that looks in the podcast normalization guide.
  4. Do not chase 0 dB. A master that peaks at exactly zero has no room for the encode. Louder is not better once a platform normalizes playback anyway, and a ceiling costs you almost nothing that any listener could ever name.

A Worked Example: The Laugh That Crackled

A comedy episode measures -16.0 LUFS — perfect — with a sample peak of -0.2 dBFS. The editor's meter stayed green all session. The true-peak reading, though, is +1.1 dBTP: every big laugh overshoots between samples.

The fix is not to redo the episode. Turn the file down by 2.2 dB and the true peak lands at -1.1 dBTP, but now loudness is -18.2 LUFS. Turn it back up and the peaks break again. This is the one case where a limiter earns its keep: a true-peak limiter holds only the laugh attacks — a few dozen moments in 40 minutes — while the gain restores the level. Final report: -16.1 LUFS, -1.0 dBTP. The laughs sound identical; the crackle is gone; and the platform's encoder now has the headroom it needs.

That trade — limiter on the peaks, gain on the whole — is exactly what a two-pass normalizer does when you set a loudness target and a true-peak ceiling together, and exactly what a plain "turn it up" cannot do.

Does Sample Rate Change True Peak?

Yes, a little — and it surprises people. The same performance exported at 44.1 kHz and at 48 kHz can show slightly different true-peak readings, because the snapshots land in different places along the wave. The wave is the same; the sampling grid is not. This is one more reason the ceiling is a ceiling with margin rather than a target to graze: if your file reads -1.0 dBTP in one tool at one sample rate, a resample downstream can nudge it up a tenth or two. Files that live at -0.1 dBTP have no such luck — any nudge is a break.

It also explains a classic confusion: a file that "passed" a sample-peak check at 48 kHz in the studio fails a true-peak check at the platform. Nothing changed about the audio. Only the ruler did.

How to Check True Peak Where You Work

You do not need new software habits, just the right meter view:

  • In a DAW: enable the loudness meter's true-peak readout (most modern DAWs include a BS.1770 meter), and watch the TP value on the master bus for the final bounce — not the channel meters.
  • On an exported file: run it through any loudness tool that reports true peak; the number to look for sits next to integrated loudness in the same report.
  • On this site: the normalizer's report shows true peak before and after, alongside LUFS and loudness range, so the ceiling check rides along with the loudness check instead of being a separate chore.

Make the check a habit at the same moment as the loudness check. They are one step pretending to be two, and doing them together takes seconds longer than doing the loudness check alone — the cheapest insurance in your whole publishing routine.

Common Mistakes

Trusting the editor's peak meter. Most DAW meters show sample peak by default. "Never went red" means the samples never clipped. It says nothing about the wave between them.

Normalizing to 0 dB "for loudness." Peaks are not loudness. A file can peak at 0 dBFS and measure -25 LUFS if the loud moments are brief — and it will still clip on encode. Loudness problems are solved with the LUFS target; peak problems with the dBTP ceiling. Different numbers, different jobs.

Fixing peaks by hand-editing the loud spots. Possible on a two-minute clip, miserable on a 40-minute episode, and pointless when a limiter does it transparently in one pass.

Checking the WAV but shipping the MP3. Encoding is one of the two peak-raising steps. If you verify true peak on the WAV and then encode carelessly — or let a platform encode a WAV that was already at 0 dBTP — the ceiling never protected the file listeners actually receive. Verify the final artifact when you can; when the platform encodes, keep your master's ceiling at -1 dBTP so their encode has room.

Assuming quiet files are safe. A quiet file has quiet peaks now. Normalize it up 8 dB to reach a loudness target and its peaks rise 8 dB too. Peak safety is checked after the gain, not before — which is why measurement happens on both passes.

True Peak in Broadcast

Broadcast rules are stricter because the chain is longer: EBU R128 caps true peak at -1 dBTP and the US ATSC A/85 practice uses -2 dBTP. If you deliver to broadcast, the EBU R128 guide covers the full spec.

The longer chain is literal: a broadcast file passes through playout servers, processors, transmitters, set-top boxes and televisions, several of which resample or re-encode. Each step can add a fraction of a decibel to peaks. Delivery specs learned this the hard way, and podcasters inherit the lesson free: the -1 dBTP ceiling exists because nobody controls the whole path from your export to a stranger's earbuds.

Troubleshooting

My report shows true peak over 0 dBTP but I hear no distortion. You may not — on your system. Distortion appears when a converter or encoder meets the overshoot, and chains differ. Listeners on other phones, cars and smart speakers run other chains. The number is the warning; fix it while it is still a number.

Holding -1 dBTP left my episode 1.5 LU under target. Your source is unusually peaky — probably loud transients over quiet speech. Soften the peaks in the mix (a fast limiter on the voice bus, or manual clips on the worst attacks), then re-normalize. Or accept it: 1.5 LU under target with clean peaks beats on-target with crackle, and repairing the underlying recording beats both.

Two meters give two different true peaks. Meters oversample by different factors and implementations vary slightly; differences of 0.1–0.3 dBTP between tools are normal. Work to the stricter reading and keep a small margin — another reason -1 dBTP, not -0.2, is the working ceiling.

Frequently Asked Questions

What does dBTP stand for? Decibels true peak: the level of the reconstructed wave's highest point, relative to full scale. dBFS readings describe samples; dBTP readings describe the wave itself.

Can true peak exceed 0? Yes — +1 or +2 dBTP readings are real. The samples stayed under full scale; the wave between them did not. Any player that reconstructs that wave faithfully is asked for a level beyond its range.

Is a limiter the only way to control true peak? Turning the whole file down also works and is cleaner — it just costs loudness. The limiter matters when you need the loudness target and the ceiling at once, which is most real episodes.

Do I need to worry about true peak for spoken word? Speech is peaky by nature — consonants and laughter are short, sharp events over a moderate average. Spoken word is exactly where sample meters lie most kindly. Yes, worry about it; it is cheap to fix and embarrassing to ship.

What ceiling should I use for YouTube and Spotify? -1 dBTP, paired with their -14 LUFS reference. The pairing is the platform guidance, and one -14 master serves both.

My loudness is fine. Can I fix true peak without touching loudness? Yes — a true-peak limiter alone, or a small overall turn-down of 1–2 dB, brings an over-peak file under the ceiling with no other change. The turn-down costs a little loudness; the limiter costs none on average. Either is better than shipping the overshoot and letting a stranger's phone decide what the distortion sounds like.

References

← All guides