ACX Audio Requirements: The Complete Technical Guide

ACX enforces eight technical requirements on every audiobook file you submit: RMS loudness between -23 and -18 dBFS, true peak below -3 dBFS, noise floor below -60 dBFS, sample rate of 44,100 Hz, mono channel, MP3 CBR at 192 kbps, 0.5–1 second of head silence, and 1–5 seconds of tail silence.[1] Files that don't meet every spec get rejected. No exceptions, no manual overrides. This guide explains what each requirement means, how ACX measures it, and how to make sure your files pass.

If you're producing audiobooks for Audible, Apple Books, or other retailers through ACX, every chapter file must clear these automated checks before a human reviewer ever hears your audio. For a closer look at both the automated and human review layers, see what ACX actually checks. The complete ACX platform page has the full spec table and what ChapterPass automates for each requirement.

What Is the ACX Standard?

The ACX standard is a set of eight measurable audio specifications that every audiobook file must pass before Audible, Amazon, or Apple Books will accept it for sale. These specifications cover loudness (RMS between -23 and -18 dBFS), peak levels (true peak below -3 dBFS), noise floor (below -60 dBFS), sample rate (44,100 Hz), channel configuration (mono), file format (MP3 at 192 kbps CBR), and silence padding at the start and end of each chapter.[1] ACX enforces these requirements through automated quality checks that reject any file failing a single specification.

What Are the 8 ACX Audio Requirements?

The ACX audio submission requirements are eight measurable technical specifications enforced on every chapter file.[1] Here's the full list with exact values:

ACX runs automated quality checks on every file you submit. Files that fail any single spec are rejected. Here's what each one means and why it matters.

1. RMS Loudness: -23 to -18 dBFS

RMS (Root Mean Square) is a measurement of average signal power across your entire chapter file. It represents how loud the chapter sounds to a listener over its full duration. ACX requires each file to land between -23 and -18 dBFS.[1] This range ensures listeners don't need to constantly adjust volume between chapters or between different audiobooks.

The ACX loudness range overlaps with the broadcast loudness standard EBU R128, which targets -23 LUFS.[2] The two standards use different measurement systems (ACX uses RMS in dBFS; EBU R128 uses LUFS with frequency weighting), but for spoken word audio, the values typically track within 1–2 dB of each other.

Most raw recordings from home studios come in around -30 to -40 dBFS. Getting into the ACX range requires gain adjustment and often light compression to keep dynamic range consistent without squashing the performance. If your file lands outside this range, it's one of the most common reasons audiobooks get rejected. For a deeper look at how ACX calculates RMS, see ACX RMS loudness explained.

Recommended target: Aim for -20 dBFS. This gives you 2 dB of margin above the -23 floor and 2 dB below the -18 ceiling.

2. True Peak: Below -3 dBFS

True peak is a measurement of the actual analog waveform reconstructed between digital samples, using 4× oversampling as defined by the ITU-R BS.1770 standard.[3] It's different from sample peak, which only looks at the discrete values your audio file stores.

How inter-sample peaks happen: Digital audio stores discrete sample points, but the real analog waveform curves smoothly between those points. When two consecutive samples are both near the maximum level, the curve between them can overshoot. This overshoot is called an inter-sample peak, and it's typically 0.5–1.5 dB higher than the sample peak. Standard peak meters miss these entirely, which is why a file can look fine on your DAW meters but fail ACX's true peak check.

ACX sets the ceiling at -3 dBFS to leave headroom for MP3 encoding, which can raise peak levels by 0.5–1 dB.[1] A brickwall limiter with true peak detection is the standard tool for ensuring compliance.

Recommended target: Set your limiter ceiling to -3.1 dBFS or lower.

3. Noise Floor: Below -60 dBFS

The noise floor is the RMS level of background sound in your recording when nobody is speaking: room tone, HVAC noise, computer fans, electrical hum, and other environmental sounds. ACX requires this to stay below -60 dBFS.[1]

Achieving a low noise floor starts with good recording conditions, but post-production tools like noise gates and noise reduction can help bring marginal recordings into spec. The key is reducing noise without introducing artefacts that sound worse than the original problem.

Watch for gain-induced noise: When you boost gain to meet RMS targets, you boost everything, including noise. A room that measured -65 dBFS noise floor before processing might measure -55 dBFS after a 10 dB gain boost. Always measure noise floor after loudness adjustment, not before.

Recommended target: Below -65 dBFS in your raw recording, giving you 5 dB of headroom for gain adjustments during mastering.

4. Sample Rate: 44,100 Hz

Sample rate is how many times per second the audio signal is measured. 44,100 Hz means 44,100 measurements per second. This is CD-quality audio (IEC 60908) and the standard for consumer distribution. It captures frequencies up to 22,050 Hz (the Nyquist frequency), well above the range of the human voice.[1]

If you recorded at 48 kHz or 96 kHz (common DAW defaults), you'll need to resample before submission. Do format conversion first, since every subsequent processing step depends on sample rate.

5. Channels: Mono

ACX requires mono audio. Spoken word doesn't benefit from stereo separation, and mono files are half the size, which is important for streaming delivery across Audible's apps and devices.[1] Other platforms like Findaway Voices accept both mono and stereo.

If your recording setup captures stereo, downmix to a single channel. Check for phase issues afterward. If stereo channels are out of phase, summing to mono can cause cancellation and thin-sounding audio.

6. Output Format: MP3, 192 kbps CBR

The final deliverable must be an MP3 file encoded at 192 kbps with a constant bit rate (CBR). Variable bit rate (VBR) MP3s will be rejected even if the average bitrate is 192 kbps, because the bitrate mode is read from the file header.[1]

Use a quality encoder (LAME is the industry standard) and explicitly set CBR mode. Some encoding tools default to VBR, so verify your settings. For a comparison across platforms, see ACX vs Findaway audio requirements. For a deeper dive into format requirements across all major distribution platforms, see the audiobook format requirements guide.

7. Head and Tail Silence

Each chapter must start with 0.5 to 1 second of silence and end with 1 to 5 seconds of silence.[1] This silence should be room tone level, not digital silence (absolute zero). A bit of ambient noise in the silence sounds natural. Dead silence followed by speech is jarring and can trigger noise floor measurement at unexpected levels.

For detailed guidance on silence padding, see audiobook silence padding explained.

8. File Naming and Consistency

While not a DSP measurement, ACX also requires consistent audio quality across all chapters. Large variations in loudness, noise floor, or tone between chapters trigger a manual review and potential rejection.

Processing all chapters through the same mastering chain ensures consistency. Batch processing is particularly useful here because it applies identical settings to every file. See the audiobook chapter formatting guide for naming conventions and structural requirements.

How Does ACX Actually Measure Your Files?

Understanding how ACX's automated checks work helps you avoid edge cases that catch producers off guard.

RMS measurement is full-file. ACX calculates RMS across the entire chapter file, including the silence padding. This means very long tail silence can pull your measured RMS lower. A chapter with generous 5-second tail silence will measure slightly quieter than the same chapter with 1.5 seconds. If your RMS is right at the -23 dBFS edge, trimming excess tail silence (while staying within the 1–5 second range) can help.

Noise floor is measured in silence only. ACX identifies the silent sections and measures their RMS. A noise gate is effective specifically because it targets the sections ACX measures. Noise during speech doesn't factor into the noise floor calculation, though it may still be flagged by a human reviewer if it's distracting.

True peak measurement uses oversampling. ACX doesn't just check sample values. They reconstruct the analog waveform between samples using 4× oversampling (176,400 points per second at 44.1 kHz) per the ITU-R BS.1770 algorithm.[3] This is why your DAW's standard peak meter might show -3.5 dBFS while ACX's check sees -2.8 dBFS. Use a true-peak-aware meter for accurate pre-submission verification.

Format checks are metadata-level. Sample rate, channel count, bitrate, and CBR/VBR are read from the file header. These checks happen before any audio analysis, which is why format rejections come back almost instantly while level rejections may take longer.

What Should You Check Before Submitting to ACX?

Use this before hitting "Submit" on ACX. Check the final MP3, not your project file or intermediate WAV. MP3 encoding changes peak levels and can shift loudness slightly.

File Format Checks

Common mistake: Many MP3 encoders default to VBR mode. Even if the average bitrate is 192 kbps, a VBR file will be rejected. Explicitly set CBR in your export settings.

Audio Level Checks

Important: Standard peak meters in most DAWs show sample peaks, not true peaks. Use a dedicated true-peak meter like Youlean Loudness Meter (free) or iZotope Insight for accurate measurement.

Content and Consistency Checks

• Opening credits file with title, author, and narrator name
• Closing credits file with end-of-book statement
• Consistent loudness across chapters (no more than 1–2 dB variation)
• No recording artefacts: clicks, pops, echo, mouth noise
• Files named and sorted in correct order

One failed spec means one rejection. Verify every chapter against every specification before submitting. For a detailed walkthrough, see how to check if your audiobook meets ACX requirements.

How Do You Meet ACX Requirements Step by Step?

This is the mastering workflow that takes your raw recording to an ACX-compliant file. Process every chapter through the same chain with identical settings.

Step 1: Convert to the Required Format

Start from your highest-quality source (WAV, not MP3). Set the project to 44,100 Hz sample rate and mono. If you recorded at 48 kHz or 96 kHz, resample first. If stereo, downmix to mono. Do format conversion before any other processing.[1]

Step 2: Apply a High-Pass Filter

Set a high-pass filter around 80 Hz with a gentle slope (12–24 dB/octave). This removes low-frequency rumble from building vibration, HVAC, and plosive energy below the useful range of speech. Technically optional for passing ACX specs, but it makes noise floor treatment more effective.

Step 3: Manage the Noise Floor

If your noise floor is above -65 dBFS in the raw recording, treat it now, before loudness adjustment. A noise gate silences pauses when levels drop below a threshold. Spectral noise reduction handles continuous background noise, but use it conservatively (6–8 dB per pass maximum) to avoid hollow-sounding artefacts.

Step 4: Adjust Loudness

This is the most critical step. Target -20 dBFS RMS, which gives comfortable margin within the -23 to -18 range. Start with gentle compression (2:1 ratio, -20 dB threshold) to even out dynamics, then apply gain to reach your target. Don't over-compress. Audiobook narration should retain natural dynamics. For detailed loudness guidance, see the audiobook loudness complete guide.

Step 5: Limit Peaks

Apply a brickwall limiter with a -3.1 dBFS true-peak ceiling. If the limiter is reducing more than 2–3 dB frequently, your loudness adjustment was too aggressive. Back off the gain. If using a sample-peak limiter (like Audacity's), set the ceiling to -3.5 dBFS to account for inter-sample peaks.

Step 6: Add Silence Padding

Add 0.5–1 second of silence at the beginning and 1–5 seconds at the end. Use room tone, not digital zero. Crossfade between silence and speech to avoid clicks or abrupt transitions.

Step 7: Export as MP3

Export as mono MP3, 192 kbps CBR, 44,100 Hz using a LAME encoder. After encoding, measure the final MP3. Encoding can shift peak levels by 0.5–1 dB. If true peak exceeds -3 dBFS in the encoded file, lower your pre-encoding limiter ceiling slightly and re-export.

Step 8: Verify Every Spec

Check the final MP3 against all eight specifications. Not the project file, not an intermediate WAV. The actual file you'll submit. For a complete verification walkthrough with free tools, see how to check audiobook ACX requirements. For the DIY mastering step-by-step guide, each step includes specific tool settings.

What Changed in ACX Requirements for 2026?

The core specifications (loudness, peaks, noise floor, format, silence) are unchanged for 2026. ACX hasn't modified these thresholds. What has changed is how strictly they're enforced:

Stricter true peak enforcement. ACX's automated checking now consistently catches inter-sample peaks that older systems occasionally missed. Files that might have squeaked through in previous years now fail reliably. Using a true-peak-aware limiter is no longer optional. It's essential.

Better consistency checking. ACX is more likely to flag audiobooks where chapter-to-chapter loudness varies significantly, even if each individual file passes. Processing all chapters through the same mastering chain is more important than ever.

Faster rejection turnaround. Quality control has gotten faster, which means you find out about rejections sooner but also need to be ready to resubmit quickly.

For the latest updates, see the complete ACX audio requirements for 2026.

Why Does ACX Enforce These Requirements?

Every spec serves a practical purpose in the listener's experience. The -23 to -18 dBFS loudness range ensures consistent volume across titles. The -3 dBFS peak ceiling prevents distortion on playback devices. The -60 dBFS noise floor keeps silence actually silent. The mono MP3 192 kbps CBR format optimises for streaming delivery across Audible's apps and devices.

The ACX loudness floor of -23 dBFS overlaps with the EBU R128 broadcast standard, which targets -23 LUFS for programme loudness.[2] While the two standards use different measurement approaches, the convergence reflects an industry consensus on what sounds comfortable for sustained listening.

Meeting these specs isn't just about passing an automated check. It's about delivering professional audio that sounds good on earbuds during a commute, on car speakers during a road trip, and on home systems during quiet evenings. The technical requirements encode decades of audio engineering best practices for the spoken word format. Before you submit, it's worth verifying your files against each spec so you catch problems before ACX does.

A note on ChapterPass: ChapterPass handles the measurable technical specs covered in this article: loudness, peaks, noise floor, format, and silence. ACX also reviews your recordings for audio quality (clicks, pops, echo, narration consistency). Those depend on your recording, not on mastering.