True peak is one of the most misunderstood specifications in audiobook production. Most narrators learn about RMS loudness and noise floor quickly; they're intuitive concepts. But true peak trips up even experienced audio engineers because it measures something you can't directly see in a standard waveform display.
ACX requires true peak below -3 dBFS on every chapter file.[1] Fail this check and your audiobook gets rejected, even if every other specification passes. This guide explains what true peak actually measures, why it differs from the peak level your DAW shows you, and how to control it reliably.
If you've already been rejected for peak issues, the ACX rejection troubleshooting guide has the specific fix workflow. For the full set of ACX specifications, see the ACX audio requirements guide.
What Is True Peak?
Digital audio represents a continuous sound wave as discrete numerical samples. At 44,100 Hz (ACX's required sample rate), there are 44,100 individual sample values per second.[1] Each sample is a snapshot of the waveform's amplitude at that exact moment.
Sample peak is the highest absolute value among all discrete samples. It's what most audio editors display in their peak meters. When Audacity's ACX Check reports "peak level," it's reporting sample peak.[2]
True peak is the estimated maximum amplitude of the continuous analog waveform that the samples represent, including points between samples. Because the original sound wave is continuous and samples are discrete, the actual waveform can pass through points higher than any individual sample.
Think of it this way: if you measure a mountain's height every 100 metres along a path, you might miss the actual summit between two measurement points. True peak estimates the summit, not just the highest point you happened to sample.
How Is True Peak Measured?
True peak measurement uses oversampling, mathematically reconstructing the waveform at a higher sample rate to estimate inter-sample peaks. ITU-R BS.1770 specifies 4× oversampling for true peak detection.[3] At 4× oversampling of 44,100 Hz audio, the measurement evaluates 176,400 points per second instead of 44,100.
Why Do True Peak and Sample Peak Differ?
The difference depends on audio content. For most speech, it's typically 0.5 to 1.5 dB, but it can be larger.[4]
When Is the Difference Small?
Slowly varying waveforms (sustained vowels, gentle transitions) have small inter-sample peak differences because the waveform doesn't change much between consecutive samples.
When Is the Difference Large?
Fast transients create larger differences. The most problematic sections in audiobook content:
- Plosive consonants (p, b, t, k sounds) producing sharp transient peaks
- Sibilants (s, sh sounds) containing high-frequency energy
- Sharp breaths between phrases
- Page turns or physical sounds that weren't fully edited out
A Practical Example
Two consecutive samples at -4.0 dBFS and -4.2 dBFS, both well below -3 dBFS. But the reconstructed waveform between them might peak at -2.7 dBFS. A sample-peak meter shows -4.0 dBFS. A true-peak meter shows -2.7 dBFS. ACX measures true peak. Your file fails.
Why Does ACX Enforce a True Peak Ceiling?
ACX sets -3 dBFS for three practical reasons:
MP3 Encoding Headroom
ACX requires MP3 at 192 kbps CBR. MP3 is a lossy codec, and its reconstruction can introduce peak levels that weren't in the original file. MP3 encoding typically raises peaks by 0.5–1 dB.[4] The -3 dBFS ceiling provides headroom for this codec-induced shift. Without it, the final audio could contain clipping, signal exceeding 0 dBFS, which causes audible distortion.
Playback Device Headroom
Audible listeners use phones, earbuds, car stereos, smart speakers. Some devices apply EQ, volume normalisation, or dynamic processing that can raise peaks. The headroom ensures the audio survives these processing stages without clipping.
Catalogue Consistency
Audible hosts hundreds of thousands of audiobooks. The -3 dBFS ceiling, combined with the RMS range, ensures baseline consistency across the catalogue.
How Do You Measure True Peak?
Which Tools Measure True Peak?
Measures true peak:
- Youlean Loudness Meter (free VST/AU plugin and standalone)[5]
- iZotope Insight / RX
- FabFilter Pro-L 2
- FFmpeg loudnorm filter
- NUGEN Audio VisLM
Measures sample peak only:
- Audacity's ACX Check plugin[2]
- Audacity's built-in peak meters
- Most DAW level meters in default mode
If you're using Audacity exclusively, you cannot directly measure true peak. Use an additional tool or leave sufficient margin. The free audiobook quality checker guide covers your options.
How Do You Measure with FFmpeg?
The most accurate free method:
ffmpeg -i your_chapter.mp3 -af loudnorm=print_format=json -f null -
This outputs JSON including input_tp, your file's true peak in dBFS. Above -3.0 means it fails ACX's check.
How Do You Measure with Youlean?
- Open Youlean Loudness Meter (standalone or as a DAW plugin)
- Load your MP3 file
- Play through the entire file
- Read the "True Peak" value
Youlean provides ITU-R BS.1770-compliant true peak measurement.[5]
How Do You Control True Peak?
True-Peak-Aware Limiting
The most reliable method. True-peak-aware limiters oversample the signal internally, detect inter-sample peaks, and apply gain reduction to keep true peak below ceiling.
True-peak-aware limiters:
- FabFilter Pro-L 2 (paid, widely considered the best)
- TDR Limiter No6 (free, supports true peak mode)
- iZotope Ozone Maximizer (paid)
- Waves L2 Ultramaximizer (paid)
Set ceiling to -3.0 dBFS (or -3.1 dBFS for margin) and the limiter handles the rest.
Sample-Peak Limiting with Margin
If using Audacity or another tool without true-peak limiting, set the sample-peak limiter lower:[2]
- For -3.0 dBFS true peak target: Set limiter to -3.5 dBFS (safe for most content)
- For extra margin: Set to -4.0 dBFS (safer but slightly reduces dynamic range)
The Audacity ACX settings guide uses the -3.5 dBFS approach.
Post-Limiting Verification
Always verify true peak after limiting and after MP3 encoding. The order matters:
- Apply limiter
- Export as MP3
- Measure true peak of the MP3 (not the WAV/project)
MP3 encoding can shift peaks upward. A file measuring -3.1 dBFS as a WAV might measure -2.8 dBFS after encoding.[4]
Where Does True Peak Fit in the Mastering Chain?
Peak limiting must happen after loudness normalisation (which can raise peaks above ceiling) and before MP3 encoding (so you can verify the encoded file):
- Format conversion (44.1 kHz, mono)
- High-pass filter (80 Hz)
- Noise treatment
- Compression
- Loudness normalisation (to -20 dBFS RMS)
- Peak limiting ← true peak controlled here
- Silence padding
- MP3 encoding
- Verification
The complete audiobook mastering guide covers the full chain.
How Do You Measure True Peak in Audacity, REAPER, and iZotope RX?
Each tool handles true peak differently. Here's exactly what each one does and doesn't measure.
Audacity
Audacity's built-in meters and the ACX Check plugin measure sample peak only, not true peak.[2] This is a known limitation. When ACX Check reports your peak level, it's reporting the highest discrete sample value, not the estimated inter-sample maximum. The two numbers can differ by 0.5 to 1.5 dB.
What this means in practice: If Audacity reports your peak at -3.2 dBFS, your true peak might be -2.0 dBFS. You would fail ACX's check despite passing Audacity's.
How to compensate in Audacity: Use Audacity's Effect > Limiter and set the limiter ceiling to -3.5 dBFS rather than -3.0 dBFS. This 0.5 dB additional margin covers the typical inter-sample overshoot for speech content.[2] For content with many plosives or sharp sibilants, -4.0 dBFS is safer.
There is no way to measure true peak directly within Audacity. After exporting your MP3, verify true peak using Youlean Loudness Meter or FFmpeg before submitting.
REAPER
REAPER can measure true peak during the render process, which is more useful than a separate measurement step. When rendering, enable the "Render statistics" option in REAPER's render dialog. After the render completes, open "Render Statistics" (under the File menu or from the render dialog) to view integrated LUFS, true peak in dBTP, and maximum sample peak for the rendered file.
REAPER's true peak measurement uses oversampling (at 2x or 4x) applied to the rendered output, which gives you an accurate post-encode figure. This means you're measuring true peak of the actual MP3 or WAV you've produced, which is what matters for ACX compliance.
What this means in practice: Set REAPER's peak limiter to -3.1 dBFS ceiling. After render, check the true peak stat. If it reads above -3.0 dBTP, lower the limiter ceiling by 0.2 dB and re-render. REAPER's rendering statistics remove the guesswork from true peak verification.
iZotope RX
iZotope RX offers true peak measurement in two places: Waveform Statistics and the Loudness Control module.
Waveform Statistics (available in most RX versions) shows true peak alongside sample peak, RMS, and other measurements. Select your audio, run Waveform Statistics, and read the "True Peak" row. This gives you a pre-encode measurement from within your RX session.
Loudness Control (RX 8 Standard and above) is the more powerful option for audiobook work. It includes a pre-built Audiobook preset that targets -20 LKFS integrated loudness and enforces a -3.0 dBFS true peak maximum. The Loudness Control tool applies a post-limiter automatically if the true peak ceiling would otherwise be exceeded, handling the full ACX loudness and peak correction in a single operation.
For audiobook producers already working in RX for noise reduction and restoration, the Loudness Control module makes it possible to handle the full mastering chain without leaving RX.
After any tool, verify the MP3. Whatever tool you use to limit peaks during export, always measure the final MP3 file (not the project WAV) with a true-peak meter. MP3 encoding shifts peak levels upward by 0.5 to 1 dB, and the only way to confirm compliance is to measure the encoded output.[4]
What Happens When a File Passes Sample Peak but Fails True Peak After Encoding?
This is the most frustrating true peak scenario: your DAW reports the file is fine, ACX rejects it.
Why does this happen?
There are two compounding factors. First, your tool measured sample peak rather than true peak, so the pre-export reading was already 0.5 to 1.5 dB lower than actual. Second, MP3 encoding raised peaks by a further 0.5 to 1 dB.[4] Together, a file that measured -3.2 dBFS sample peak before encoding might measure -1.7 dBFS true peak after encoding. That's a 1.5 dB compound error, enough to cause a clear failure.
The fix
Work backward from the required output. ACX checks the final MP3. So measure the final MP3.
Step 1: Export your current file as MP3 (192 kbps CBR, 44,100 Hz, mono).
Step 2: Measure true peak of the MP3 using Youlean or FFmpeg. Note the actual true peak value.
Step 3: Determine the gap. If your MP3 true peak is -1.7 dBFS and your target is -3.0 dBFS, you need 1.3 dB more headroom.
Step 4: Go back to your pre-export project file. Lower your limiter ceiling by the gap amount plus 0.2 dB margin. In this example, lower from -3.1 dBFS to -4.6 dBFS.
Step 5: Re-export as MP3 and re-measure. Confirm the MP3 true peak is now below -3.0 dBFS.
Prevention
Use a true-peak-aware limiter at -3.1 dBFS. This accounts for inter-sample peaks before encoding. After encoding, verify the MP3 true peak. If it's above -3.0 dBFS, lower the pre-export ceiling by 0.3 to 0.5 dB and re-export. For most content, -3.3 to -3.5 dBFS pre-export ceiling produces a final MP3 that passes comfortably.[5]
When specific sections keep failing
If true peak failures are concentrated in specific passages rather than spread throughout the file, the cause is almost always transient content: plosive consonants, sharp sibilants, or physical sounds (desk movement, page turns) that weren't fully edited out.
For plosives, a de-esser or plosive reduction step before the limiter reduces the transient energy that causes inter-sample peaks. For physical sounds, find and cut them in editing before mastering. A limiter with lookahead (standard in most true-peak-aware limiters) handles fast transients better than limiters without lookahead, but it cannot compensate for transients that are many dB above the target. Fixing extreme transients at the editing stage is more reliable than asking the limiter to do all the work.
What Are Common True Peak Problems?
Sample Peak Passes, True Peak Fails
Cause: Using a sample-peak meter or limiter without enough margin. Fix: Use a true-peak-aware limiter, or lower the sample-peak ceiling to -3.5 dBFS or lower.
True Peak Passes as WAV, Fails as MP3
Cause: MP3 encoding raised peaks by 0.5–1 dB. Fix: Lower limiter ceiling (try -3.5 if you were at -3.1). Always verify the final MP3.
Failures on Specific Sections Only
Cause: Transient content (plosives, sharp breaths) creating inter-sample peaks above ceiling. Fix: Ensure limiter lookahead is enabled. Most true-peak limiters use it to catch fast transients.
Aggressive Limiting Causing Distortion
Cause: Pre-limiter peaks too far above ceiling, making the limiter work too hard. Fix: Add more compression upstream. A 2:1 compressor brings peaks closer to the average, reducing limiter workload.