Introduction: Balancing Clarity and Authenticity in Audiobook Production

Producing a professional audiobook requires more than a compelling voice and a solid recording setup. The final audio must meet strict technical standards — especially those set by ACX (Audiobook Creation Exchange) — while preserving the natural warmth and presence of the narrator’s performance. Noise reduction is one of the most critical — and most misused — tools in the post-production chain. When applied incorrectly, it can strip the life out of a recording, introduce digital artifacts, or fail to meet ACX’s noise‑floor requirements. This guide walks you through best practices for using noise reduction effectively, ensuring your audio passes ACX quality checks without sounding overprocessed or artificial.

Understanding ACX Audio Specifications

Before reaching for any noise reduction plugin, you must understand the technical boundaries ACX imposes. The platform specifies a maximum noise floor of -60 dB (A‑weighted), a RMS level between -23 dB and -18 dB, and peak levels no higher than -3 dB. These limits exist to guarantee consistent playback across devices and to avoid listener fatigue. Noise reduction that pushes the noise floor below -60 dB is often acceptable, but only if it does not degrade the signal or introduce audible anomalies.

ACX also evaluates recordings subjectively. Overly processed audio — even if technically within spec — may be flagged during human review. This is why a light touch is essential: your goal is to reduce background noise while preserving the original character of the voice, including its transient details, breath sounds, and natural tonal balance.

What Is the Right Amount of Noise Reduction?

Many narrators fall into the trap of believing that more noise reduction automatically yields a cleaner recording. In reality, excessive processing can make the voice sound hollow, “tinny,” or muffled — especially when the noise floor is pushed too far below -70 dB. The sweet spot for narration generally lies in reducing the noise floor to somewhere between -62 dB and -70 dB, depending on the original recording’s cleanliness.

If your raw recording already has a noise floor around -50 dB or higher, you may need to apply multiple gentle passes rather than one aggressive reduction. Each pass should remove only 6–10 dB of noise, with careful listening afterward. This approach minimizes the risk of audible artifacts, such as “musical noise” or a warbling quality in sibilant sounds.

Common Types of Background Noise

  • Constant low-level hum – often from computer fans, HVAC systems, or electrical interference. Narrowband filters or spectral tools work well.
  • Rumble – low-frequency noise from traffic, footsteps, or building vibrations. A high-pass filter (80–100 Hz) can often remove this without affecting voice.
  • Buzzing or high-frequency hiss – from noisy preamps, poor cables, or room reflections. Broadband noise reduction plugins are the typical solution.
  • Impulsive noises – clicks, pops, chair squeaks. These are best removed manually with spectral editing rather than with traditional noise reduction.

Identifying the type of noise you’re dealing with helps you choose the right tool and avoid unnecessary processing.

Selecting the Right Noise Reduction Tools

Not all noise reduction software is created equal. Low‑cost or built‑in tools may offer limited control and cause more harm than good. For audiobook work, invest in professional‑grade tools that allow you to adjust parameters like reduction amount, noise floor target, and spectral sensitivity.

Top Tools for ACX‑Compliant Audiobook Production

  • iZotope RX – Industry standard for audio restoration. The Voice De-noise module is specifically designed for speech, offering a Learn function to sample noise and apply reduction with adjustable quality and artifact suppression.
  • Adobe Audition – Its adaptive noise reduction effect provides real‑time preview and a range of presets. The “Noise Reduction (process)” effect works well for removing constant background noise with minimal artifacts.
  • Waves NS1 and Clarity Vx – Lightweight plugins that apply real‑time noise reduction. They’re particularly useful for live monitoring during recording to reduce the need for heavy post‑production.
  • Cedar DNS One – A higher‑end option often used in broadcast and post‑production for transparent noise suppression without degrading voice quality.

Whichever tool you choose, avoid cheap “de‑noiser” VST plugins that promise one‑click perfection. Always test each tool on a short sample before committing to a full passage.

Best Practices for Applying Noise Reduction

Following a consistent workflow will help you stay within ACX guidelines while preserving natural sound.

1. Record the Cleanest Audio Possible

Noise reduction should never be a crutch. The better your raw recording, the less processing you’ll need. Use absorption panels, a well‑isolated room, and proper mic technique. Monitor your levels and check for any constant background noise before you start a long session.

2. Sample the Noise Profile Correctly

Most noise reduction tools require a noise print — a short sample of background noise taken from a silent moment. Capture 2–3 seconds of pure room tone (the background noise without speech). If your recording contains movement or intermittent sounds, manually isolate a clean section of silence (e.g., at the beginning or end of a chapter).

3. Apply in Small Increments

Start with a reduction of 12–18 dB for moderate background noise, then listen critically. Increase or decrease in steps of 3–6 dB. Some tools also let you adjust a “reduction amount” vs. “noise floor” parameter. Keep the noise floor target at or slightly below -60 dB but not below -75 dB, as that often introduces artifacts.

4. Use Spectral Editing for Problem Spots

If a single page turn or breath creates an unwanted sound, avoid broad noise reduction. Instead, use a spectral editor (like iZotope RX’s Spectral Repair) to remove the isolated noise. This prevents affecting adjacent clean audio.

5. Always Compare Before and After

Toggle the effect on and off while listening through good headphones. Pay attention to changes in the voice: does it sound thinner? Are there any high‑frequency artifacts? Does the sibilance become harsh? If you hear any degradation, reduce the strength or adjust the frequency range of the noise reduction.

6. Check ACX Compliance After Processing

After noise reduction, run your audio through an ACX‑compliance checker (many DAWs offer this, or use a free tool like ACX’s own Audio Submission Guidelines). Verify RMS, peak level, and noise floor. If the noise floor is above -60 dB, you may need to apply a slight additional reduction, but only if the voice still sounds natural.

Common Mistakes to Avoid

Even experienced producers sometimes stumble when using noise reduction. Here are pitfalls specifically relevant to ACX audiobooks:

  • Applying noise reduction to the entire track without auditioning. Always listen to a representative sample — if you process an entire two‑hour file and then discover artifacts, you’ll have to redo the whole pass.
  • Over‑reducing the noise floor below -80 dB. This often makes the voice sound “distant” or “thin” because the natural room ambience is removed. ACX prefers a small amount of ambient room tone over a dead, processed sound.
  • Using noise reduction to fix poor recording technique. Heavy processing cannot fix clipping, plosives, or severe room echo. These issues require re‑recording or dedicated tools (de‑esser, de‑click, de‑reverb).
  • Mixing noise reduction with compression before checking levels. Apply noise reduction first, then adjust levels with compression or limiting. This order prevents the compressor from amplifying residual noise.

Noise Reduction and the Narrator’s Performance

A great audiobook relies on the narrator’s connection with the listener. Over‑processed audio can break that connection. When listeners feel the sound is artificial or “soulless,” they are more likely to stop listening or leave negative reviews. Therefore, your noise reduction strategy should always prioritize preserving the voice’s natural dynamics. This means leaving some room tone intact — especially for pauses and breaths — so that the audio doesn’t sound flat or robotic.

Some narrators even prefer to avoid noise reduction entirely if the original recording already has a noise floor of -55 dB or better. In those cases, a light high‑pass filter at 80 Hz and a subtle broadband expansion can be enough to meet ACX specs without harming the performance.

Step‑by‑Step Noise Reduction Workflow for ACX

Step 1: Assess Your Raw Recording

Import a chapter into your DAW. Open the spectrum analyzer or a metering plugin to check the noise floor during silence. If it’s already at -55 dB or lower, you may skip aggressive noise reduction. If it’s higher, proceed to step 2.

Step 2: Capture a Noise Print

Highlight a 2‑second section of pure room tone (no speech, no movement). Use your tool’s “Learn” or “Capture” function to sample this noise profile.

Step 3: Apply Gentle Reduction

Start with a reduction level of 12 dB. Process a short test segment (10–15 seconds) that includes both silence and dialogue. Listen carefully: Does the voice sound natural? Are there any artifacts on sibilants or fricatives? If not, continue. If yes, reduce the amount.

Step 4: Check the Noise Floor

After processing, check the noise floor again. If it’s still above -60 dB, increase reduction by 3‑6 dB. If it’s below -60 dB but the voice sounds fine, you’re good. If below -65 dB and the voice sounds degraded, undo and try a different approach (e.g., a separate high‑pass filter for low‑end rumble).

Step 5: Listen at Low Volume

Play the processed segment at a low listening level (e.g., 40 dB). At this level, artifacts like “warbling” or “musical noise” become more apparent. If you hear anything unnatural, adjust the reduction amount or use a tool with better artifact suppression.

Step 6: Apply to the Full File

Once you’re satisfied with the test segment, apply the noise reduction to the entire chapter. Use offline processing (non‑real‑time) if your tool offers it, as real‑time processing can introduce variations.

Step 7: Final Quality Control

Run the chapter through an ACX compliance checker. Verify RMS, peak, and noise floor. Then do a full listen — the most important test is one complete pass through the audio while reading along with the text. Mark any spots where the voice sounds unnatural or where noise reduction has created an audible “pumping” effect.

Advanced Techniques: Spectral Repair and Adaptive Noise Reduction

For narrators dealing with complex noise (e.g., intermittent traffic, air conditioning cycles, or distant conversations), adaptive noise reduction tools can be more effective than static profiles. iZotope RX’s “Voice De‑noise” includes an adaptive mode that continuously updates the noise estimate, allowing it to respond to changes in background noise over time. Similarly, Adobe Audition’s “Adaptive Noise Reduction” effect applies real‑time spectral adjustments.

These tools require careful threshold settings. If you set the adaptation too fast, the voice can temporarily be reduced during pauses; too slow, and the noise may bleed through. A good starting point is to set the adaptation speed to “Medium” and the noise reduction to 12–15 dB, then listen and adjust.

Conclusion

Mastering noise reduction for ACX audiobooks is not about achieving absolute silence; it’s about achieving a natural, comfortable listening experience that meets technical requirements. By recording cleanly, choosing professional tools, applying processing in small increments, and always listening critically, you can eliminate background noise without sacrificing the warmth and presence of your narration. Remember: the best noise reduction is the one your listeners never notice. For further guidance, refer to the official ACX Audio Submission Guidelines and explore the restoration resources at iZotope’s audiobook tutorial page.