Every streaming loudness guide tells you to hit −14 LUFS. Fewer tell you about the other number that can still ruin your file after upload: true peak. This page explains what it is in plain terms, why a meter reading of "0 dB" does not mean "safe," and which ceiling you should actually deliver — with a free way to measure both numbers in your browser. If LUFS is the average question (how loud over time), true peak is the instant one (the single highest excursion between samples).
Sample peak vs inter-sample peak: the 30-second version
Digital audio is a sample stream — at 48 kHz, you are taking 48,000 amplitude readings per second. A sample peak meter reads only those points. But the actual waveform continues between them. Two adjacent samples can both sit at −1 dBFS while the wave between them bows upward and crosses 0 dBFS — an inter-sample (IS) peak. Your DAW's meter, watching only samples, will happily show "no clipping."
This is not a theoretical curiosity:
- It happens constantly in hard-limited material (modern EDM, trap 808s, anything pushed for loudness).
- It gets worse when your file passes through an encoder: AAC and MP3 are approximations of the waveform, and that approximation can create IS peaks that were never in your master — typically adding up to about 1 dB of peak level on top.
- So a master measuring exactly 0 dBFS sample-peak can arrive at streaming as ~+1 dBTP: real clipping, audible as crackle, mostly on phone speakers where the DAC is least forgiving.
A true peak meter solves this by oversampling — internally multiplying your rate (4× or more) and reading the reconstructed wave between original samples, reporting in dBTP (decibels true-peak). At 192 kHz internal math, that is ~192,000 readings per second instead of 48,000 — enough resolution to catch what sample meters structurally cannot.
The numbers you should deliver in 2026
| Platform context | Ceiling guidance (as published/reported) | Practical rule for your master |
|---|---|---|
| EBU R 128 / AES TD1008 (the standards streaming is built on) | −1 dBTP maximum | Treat −1.0 dBTP as the outer boundary, not a target to graze |
| Apple Music / Apple Digital Masters | −1 dBTP required (validated with their afclip tooling) | Deliver at or below; -14 LUFS integrated is the common companion target |
| Spotify | Keep true peak below −1 dBFS for lossy delivery; −2 dBTP if your master is louder than −14 LUFS — a harder-limited file leaves encoders less room | Aim ≤ −1.0 dBTP at ≈ −14 LUFS; aim closer to −2.0 dBTP if you deliberately mastered hot (e.g., club cuts) |
| YouTube, Tidal and most other major platforms | −1 dBTP recommended | Same rule of thumb holds everywhere |
Two honest caveats: platform docs shift between updates, so re-verify before a big release; and "required" in practice means "you will be normalized/limited after the fact or rejected from certain delivery programs," not that your file is deleted. The cost of ignoring it is quality at the worst possible moment — post-upload, on someone else's encoder.
Where true peak actually bites (and where it doesn't)
It matters a lot when:
- Your master is loud and hard-limited: the louder you push, the closer peaks sit to ceiling, and 1 dB of encoder-added headroom hunger becomes audible clipping. This is exactly why Spotify's guidance tightens to −2 dBTP for >−14 LUFS masters.
- You deliver AI-generated output: generators frequently produce spiky material with transient-heavy top end that sample meters under-report — measure every export before you trust it (the full workflow is in mastering Suno output); the same exports often hide tempo/key surprises too, which the free BPM & key analyzer reads right from the file.
- You master for club/streaming hybrids and re-use one file across contexts.
It matters less when:
- Your material stays dynamic (acoustic, jazz, lo-fi at −16 to −18 LUFS with natural peaks): there is already headroom, encoders add 1 dB into space you don't use, and the loudness target does most of the work. What -16 vs -14 LUFS actually changes covers that side.
- You deliver a "vinyl-ready" or archival file no streaming encoder will touch (then sample peak + taste is your domain).
How to measure both numbers in two minutes
You do not need a plugin license for this:
- Open the free browser LUFS meter & normalizer — drop any MP3/WAV/FLAC. It computes integrated LUFS, true peak (dBTP) and sample peak locally on your device; nothing is uploaded anywhere.
- Read both numbers against the table above: e.g., "−13.8 LUFS, −0.4 dBTP" = loud but safe at target; "−9.2 LUFS, +0.1 dBTP" = will be turned down and may clip after encoding — a real master needed, not just a gain change (gain cannot fix peaks that are already over).
- If you only need loudness correction and your peaks have room, the same tool normalizes to −14/−9/−16 LUFS with −1 dBTP headroom built in. That is a legitimate basic fix — but it is gain+limit only; tonal shaping still requires a full master (what that adds).
A common confusion: true peak vs loudness are independent dials
"Making it louder" (LUFS) and "how high the peaks go" (dBTP) move together only while you have headroom left. Once your limiter engages to hit a LUFS target, every extra dB of loudness costs peak ceiling — that trade-off is why there is no single "perfect number." The deliverable is a pair: roughly −14 LUFS and ≤ −1 dBTP (−2 if you went hot). Mastering services exist mostly to manage that pair automatically while doing the tonal work; free tools exist so you can verify either side.
Frequently Asked Questions
What is true peak, in one sentence?
The highest point of the reconstructed waveform between digital samples (dBTP) — not just the sampled points your DAW meter displays.
Why do encoders make peaks higher than my file had?
AAC/MP3 encode by approximating the wave mathematically; those approximations can overshoot the original signal's envelope, typically adding up to ~1 dB of intersample peaks. Headroom (staying below −1 dBTP) is what absorbs that overshoot without clipping.
My DAW says 0.0 dBFS and no red light — am I safe?
No, not necessarily. Sample-peak meters structurally miss inter-sample excursions. Re-check with a true-peak meter (the free browser tool does this in-browser) before you deliver.
Should I limit to −1 dBTP or −2 dBTP?
At the standard ≈−14 LUFS streaming target, ≤ −1.0 dBTP is correct and matches Apple's requirement; if your master is deliberately louder than −14 LUFS (club/EDM), follow Spotify's own tighter guidance toward −2.0 dBTP so encoders have room — see what the loudness numbers mean for when hot masters are worth it at all.
Does true peak matter for AI-generated songs from Suno or similar?
Especially yes: generated transients often under-report on sample meters, and a spiky export that looks "fine" in your DAW can clip after DistroKid's or Spotify's encoding — measure every export (free tool above) before upload; the whole pre-delivery checklist lives at mastering before DistroKid.
Can I hear true-peak clipping?
Usually on phone speakers: short crackle/glitches on transients that desktop monitors smooth over. If a file sounds clean everywhere but crackles on one friend's phone, this is almost always the cause — re-measure before re-uploading anything.