Understanding ACX Loudness Requirements

Before diving into leveling and normalization, it's critical to understand what ACX actually demands. Amazon's Audiobook Creation Exchange (ACX) enforces specific technical standards to ensure a consistent listening experience across all devices and environments. The two primary measurements are loudness and peak level.

ACX requires an average loudness between -23 LUFS and -18 LUFS (Loudness Units relative to Full Scale), with the ideal target being -23 LUFS. Peak level must not exceed -3 dBFS (decibels relative to Full Scale). Additionally, the audio should have a noise floor no higher than -60 dBFS. These parameters are designed to produce audiobooks that don't require listeners to constantly adjust their volume, whether they are using headphones, car speakers, or a smart speaker.

If your audiobook fails to meet these specifications, it will be rejected during ACX's quality review. Rejection not only delays your release but can also lead to additional fees if you need to re-record or outsource post-production. Understanding leveling and normalization as distinct tools will help you pass ACX review on the first attempt.

What Is Leveling?

Leveling is the process of smoothing out volume inconsistencies within a single recording. A narrator's voice might naturally vary in loudness due to breath, emphasis, or changes in energy. Leveling uses dynamic range compression to reduce the gap between the quietest and loudest parts of the audio.

For audiobooks, leveling is especially important because the narration must remain intelligible and comfortable to hear over long periods. Without proper leveling, a listener might miss a whispered sentence or be startled by a sudden rise in volume. Leveling creates a more polished, professional sound that meets ACX's requirement for consistent loudness.

How to Level Your Audio in Practice

  1. Open your audio in a digital audio workstation (DAW). Popular choices include Audacity (free), Adobe Audition, Reaper, or GarageBand. Each offers a compressor or leveling tool.
  2. Identify problem sections. Scroll through the waveform and look for high peaks and deep valleys. You can also listen to the track to spot areas that are too loud or too quiet.
  3. Apply a compressor. Set a threshold around -18 dB to -12 dB, a ratio of 3:1 to 4:1, and a moderate attack (5–15 ms) and release (50–150 ms). The compressor will attenuate sounds that exceed the threshold, reducing peaks. Make sure the makeup gain is set to bring the average level back up.
  4. Use a limiter as a safety net. Follow the compressor with a limiter set to -3 dBFS. This ensures no peak exceeds the ACX maximum, even if the compressor fails to catch sudden spikes.
  5. Apply light EQ to balance the tone. Sometimes leveling can introduce a dull sound; a gentle high-shelf boost around 4–8 kHz can restore clarity without raising noise.
  6. Listen critically. Leveling should be subtle. Overcompression will make the audio sound flat, lifeless, and unnatural. ACX reviewers will reject audio that exhibits pumping, breathing, or audible compression artifacts.

What Is Normalization?

Normalization is a different beast: it adjusts the overall gain of a file so that its peak loudness hits a target level. Unlike leveling, normalization applies a uniform adjustment across the entire recording. It does not change the dynamic range—the quiet parts remain relatively quiet, and the loud parts remain loud relative to each other.

In the context of ACX, normalization is used to bring the average RMS (Root Mean Square) loudness or integrated LUFS value into the acceptable range. Many audio editors include a "loudness normalization" tool that automatically calculates the current level and applies the necessary gain to reach a target like -23 LUFS.

How to Normalize Your Audio Correctly

  1. Work on a leveled file. Normalization works best after you have already leveled the recording. If you normalize first, the compressor will later crush peaks, potentially pushing the average loudness above the target.
  2. Use loudness normalization, not peak normalization. Peak normalization only sets the highest peak to a specific value (e.g., -3 dBFS) and doesn't account for perceived loudness. LUFS normalization is much more reliable for audiobooks. In Audacity, use Effect > Loudness Normalization > Target Integrated Loudness of -23 LUFS.
  3. Apply multiple passes if needed. After the first normalization, check the integrated loudness using a meter (free tools like Orban Loudness Meter or the built-in meter in your DAW). If it's still too quiet or too loud, adjust your target or re-level before normalizing again.
  4. Always preserve headroom. ACX allows peaks at -3 dBFS, but it's safer to aim for peaks around -6 dBFS to leave a margin for error. A limiter can be used after normalization to catch any overs.

The Real Difference: Leveling vs. Normalization

Many newcomers confuse these two processes or think one can replace the other. Here's a clear breakdown:

  • Leveling (compression) reduces the dynamic range by turning down loud parts and turning up quiet parts. The overall perceived loudness can increase, but the peaks are controlled.
  • Normalization applies a single, constant gain change to the entire file. It does not alter the dynamic range; it merely shifts the entire waveform up or down.

For ACX compliance, you need both. Leveling first ensures that the dynamic range is not excessive. Normalization then adjusts the average level to hit the -23 LUFS target. Skipping either step will result in audio that is either too inconsistent or too quiet/loud.

Using the ACX Checker Tool

ACX provides a free analysis tool called the ACX Checker (available as a plugin for various DAWs or as a standalone program). You should run your finished file through this tool before uploading. It will report pass/fail for RMS loudness, peak level, noise floor, and other specs.

If the checker indicates failure, go back to your session. Common issues:

  • RMS too low: The average loudness is below -23 LUFS. Apply a gain boost via normalization (target -23 LUFS) and re-check.
  • RMS too high: Average loudness exceeds -18 LUFS. Reduce the overall level via normalization (target -23 LUFS) and ensure your compressor didn't over-apply makeup gain.
  • Peaks exceed -3 dBFS: Your limiter or compressor didn't catch them. Either tighten the limiter threshold or manually reduce those specific peaks.
  • Noise floor too high: Background noise (hiss, hum, room tone) is above -60 dBFS. Apply a noise gate or use noise reduction before leveling.

Best Practices for Meeting ACX Standards

Seasoned audiobook engineers follow a predictable workflow. Here is a production-ready sequence:

  1. Record in a consistent environment with proper microphone technique. Avoid plosives, sibilance, and mouth clicks.
  2. Edit out errors and breaths that are too loud or distracting. Use spectral analysis to remove unwanted clicks.
  3. Apply gentle compression (leveling) with a ratio of 2.5:1 to 4:1. Keep the threshold moderate to avoid artifacts.
  4. EQ subtlety to correct any tonal imbalances.
  5. Normalize to -23 LUFS using integrated loudness normalization.
  6. Apply a brickwall limiter set to -3 dBFS with a fast attack.
  7. Export a 24-bit WAV file (or 16-bit if required by ACX).
  8. Run ACX Checker and review the results. If it passes, you're ready to upload.

Additional Tips from the Field

  • Avoid excessive compression that makes your voice sound squashed. Listen in multiple environments (car, headphones, speakers) to ensure the audio translates well.
  • Monitor your audio using a true-peak meter. ACX checks true-peak values, not just sample peaks.
  • If you work with multiple chapters, level and normalize each chapter individually. Consistency across chapters is even more important than within a single file. Use reference tracks to match loudness.
  • Consider using a voice-specific compressor plugin designed for speech, such as Vocal Rider (Waves) or the classic LA-2A emulation. These can produce more natural results than a generic compressor.

Common Mistakes and How to Avoid Them

Mistake 1: Normalizing Before Leveling

If you normalize first, you set the overall level based on the raw, unleveled file. Then when you compress, the makeup gain may push the average loudness far beyond the target, or the compression may introduce artifacts because the peaks are already at the limit.

Solution: Always level first, then normalize.

Mistake 2: Over-Reliance on Peak Normalization

Peak normalization (e.g., "Normalize to -3 dBFS") does nothing to control loudness consistency. Your file might have peaks at -3 dBFS but the average loudness could be -10 LUFS, completely failing ACX.

Solution: Use loudness normalization (LUFS or RMS) and ignore peak normalization.

Mistake 3: Ignoring the Noise Floor

Even if your loudness and peaks are perfect, excessive background noise will cause rejection. Common culprits: computer fans, HVAC hum, clothing rustle, or poor microphone technique.

Solution: Apply noise reduction gently before any other processing. Use a high-pass filter to remove rumble below 80 Hz.

Mistake 4: Applying the Same Settings to Every File

Your voice and recording environment are unique. A compressor preset that worked on one narration may not work on another. Trust your ears.

Solution: Learn to listen for artifacts. Take notes on which settings work for your specific setup.

Advanced Techniques for Consistent Loudness

Once you have mastered basic leveling and normalization, you can refine your workflow further:

  • Manual volume automation: In your DAW, draw volume envelopes to gently raise or lower specific sections before compression. This pre-leveling reduces the workload on the compressor and yields a more natural sound.
  • Multiband compression: Some engineers use a multiband compressor to independently control different frequency ranges. For example, you can compress the low-mid range (where plosives and boomy resonances live) more aggressively than the upper frequencies, preserving clarity.
  • Loudness matching across chapters: After processing all chapters, use a loudness meter to compare the integrated LUFS of each. If chapter 3 is -24 LUFS and chapter 4 is -22 LUFS, adjust the gain on one chapter to make them uniform.

For a deeper dive into ACX technical specifications, refer to the official ACX audio quality requirements. You can also explore Audacity's loudness normalization documentation for step-by-step instructions. If you want to understand LUFS more thoroughly, read EBU R128, the industry standard for loudness measurement.

Final Thoughts

Leveling and normalization are complementary processes that, when applied correctly, ensure your audiobook meets ACX volume standards. Leveling tames dynamic inconsistencies, while normalization brings the overall loudness to the target. Together they produce a listening experience that is comfortable, professional, and compliant.

Invest time in learning your DAW's tools and calibrating your ears to recognize the subtle differences between adequate and excellent audio. With practice, you will be able to produce audiobooks that not only pass ACX review but also receive positive listener feedback for their polished sound.

Remember always to check the latest ACX guidelines before publishing, as standards can evolve. Master these techniques, and you'll have a smooth path to audiobook distribution on Amazon, Audible, and iTunes.