Your audiobook's noise floor must stay below -60 dBFS to pass ACX's technical review.[1] That number represents the threshold below which background noise becomes inaudible to a listener wearing earbuds in a quiet room. Above -60 dBFS, your pauses have an audible hiss, hum, or rumble that distracts from the narration. This guide covers everything: what causes noise floor problems, how to measure accurately, how to fix it through room treatment, recording technique, and post-production, plus how mastering gain interacts with noise to create the most common cascading failure in audiobook production.
What Is Noise Floor?
Noise floor is the level of background sound present in your recording when nobody is speaking.[1] It's the sum of everything your microphone picks up that isn't intentional: HVAC systems, computer fans, electrical interference, outdoor traffic, refrigerator compressors, and the ambient noise of your recording space.
Every recording has a noise floor. Even a professional studio with acoustic treatment has a measurable one. ACX measures it as the RMS level of silent sections in your final mastered MP3. The threshold is -60 dBFS. Above -55 dBFS, trained listeners can detect hiss on earbuds. Above -50 dBFS, background noise is clearly distracting.
Critical: The standard applies to the final mastered MP3, not your raw recording. Mastering raises your noise floor. If you start at -62 dBFS and add 8 dB of gain for loudness, your noise floor becomes -54 dBFS, which fails.
What Types of Noise Affect Audiobook Recordings?
Broadband noise (hiss): Covers a wide frequency range and sounds like a constant "shhh." Common sources include HVAC, computer fans, and general room ambience. Responds well to spectral noise reduction.
Tonal noise (hum and buzz): Concentrated at specific frequencies, typically 50 Hz or 60 Hz from electrical mains and their harmonics.[2] Identifiable by its consistent pitch. Responds well to notch filters.
Intermittent noise: Traffic rumble, aircraft, household activity. Can't be addressed with constant processing. Requires re-recording affected sections or careful manual editing.
Self-noise: Every microphone generates internal electronic noise. Condenser microphones typically have lower self-noise; dynamic microphones have better rejection of room noise. The trade-off depends on your room.
How Do You Measure Noise Floor Correctly?
Incorrect measurement is a frequent source of confusion and failed submissions.
Step 1: Work with your final MP3. Measure the actual file you'll submit to ACX, not the project file, not an intermediate WAV. MP3 encoding can affect noise floor measurement.
Step 2: Find representative silence. Select a section with no speech, at least 1 second, ideally 2–3 seconds. Use a natural pause between sentences in the body of the narration. Don't use artificial silence, the very start/end of the file, or sections right after loud passages where processing may still be releasing.
Step 3: Measure RMS of the selection.
- Audacity: Select the silence → Analyze → Contrast → Measure "Background" → Read RMS
- Adobe Audition: Select → Window → Amplitude Statistics → Total RMS
- iZotope RX: Select → Read RMS in the level meter
Step 4: Check multiple locations. Noise floor can vary within a chapter. HVAC may cycle, a truck may pass. Measure at least three different silence points and use the worst (highest) value.
What Causes Noise Floor to Fail After Mastering?
The primary cause is gain amplification, the most common cascading failure in audiobook production. If you've been rejected specifically for noise floor, the noise floor rejection guide covers the most direct fixes.
The formula: Final noise floor = Raw noise floor − Applied gain (in dB)
If your raw recording has a -63 dBFS noise floor and you apply 6 dB of gain to reach your RMS target, the noise floor becomes -57 dBFS, above the limit.[1] Plan your gain budget: if you need 8 dB of gain for loudness, your raw noise floor must be below -68 dBFS.
Compression effects: During speech, compression reduces peaks so less gain is needed (helps). But some compressors apply makeup gain to silent sections, raising the noise floor (hurts). Set the compressor threshold above your noise floor to avoid compressing silence.
| Environment | Typical Raw Floor | Feasible After Mastering? |
|---|---|---|
| Professional studio | -75 to -80 dBFS | Easily, generous margin |
| Well-treated home studio | -65 to -72 dBFS | Yes, comfortable margin |
| Untreated quiet room | -58 to -65 dBFS | Marginal, needs treatment |
| Average home office | -50 to -58 dBFS | Challenging, significant treatment |
| Noisy environment | Above -50 dBFS | Not feasible without re-recording |
How Do You Reduce Noise Floor Through Room Treatment?
The most effective noise reduction happens before you press record. Treating your recording environment eliminates noise at the source, which is always cleaner than removing it in post-production.
Eliminate Active Noise Sources
HVAC: The single biggest noise source in most home recordings. Turn off heating and air conditioning during recording. Plan shorter sessions with breaks if the space gets uncomfortable.
Computers: Move the computer to another room if possible (route cables through the wall). Failing that, minimise load during recording. Close unnecessary applications and prevent antivirus scans.
Appliances: Unplug the refrigerator during recording sessions. Don't run dishwashers or washing machines.
Reduce Environmental Noise
Close all windows and doors. Hang heavy curtains over windows. Apply weather stripping to seal gaps. Record in an interior room with no exterior walls if possible. In multi-level buildings, record during quiet hours to avoid structural noise transmission.
Add Acoustic Treatment
First reflection points are the most impactful treatment positions: the surfaces where sound bounces once before reaching the mic. Place 2-inch thick acoustic foam or mineral wool panels on the wall behind your monitor, the ceiling above your position, and the desk surface below the mic.
Behind the microphone: A cardioid mic picks up sound from the front and rejects the rear. Absorption behind the mic reduces room sound in its most sensitive pickup direction.
Portable solutions: Reflection filters provide moderate reduction. Heavy moving blankets draped around your recording position create a surprisingly effective temporary booth. A walk-in closet full of clothes provides excellent absorption for spoken word.
How Do You Optimise Recording Technique for Lower Noise Floor?
Get close to the mic. A shorter mouth-to-microphone distance gives more voice signal relative to room noise. The sweet spot is 6–8 inches with a pop filter.[2]
Set input gain properly. Aim for normal narration peaks at -12 to -6 dBFS on the recording meter. This gives good signal-to-noise ratio without risking clipping. Many narrators set gain too low for safety, producing recordings that need massive amplification, which amplifies every bit of noise.
Choose the right microphone for your room. In a quiet room, a large-diaphragm condenser (lower self-noise, less preamp gain needed) is usually better. In a noisy room, a dynamic mic (better room noise rejection) may be the better trade-off.
Monitor before each session. Record 10 seconds of silence and measure the RMS. If your room tone is above -60 dBFS before you've spoken, no amount of post-production will reliably fix it. Go back and address the noise source. The audiobook production workflow guide covers how recording environment decisions fit into the broader production pipeline.
How Do You Fix Noise Floor in Post-Production?
When room treatment isn't sufficient, post-production tools can bridge the gap. The key principle: apply noise treatment before loudness gain.
Step 1: High-Pass Filter (Always Apply)
Apply at 80 Hz with a 12 dB/octave slope. Low-frequency rumble from building vibration, traffic, and air pressure changes inflates your noise floor without affecting speech intelligibility. Removing it is always beneficial and introduces no artefacts.
Step 2: Notch Filters for Tonal Noise
If your recording has a consistent hum or buzz, identify the frequency with a spectrum analyser (usually 50 Hz or 60 Hz mains frequency and harmonics). Apply a narrow notch filter (Q of 10–20, cut by 10–15 dB) at each offending frequency. Notch filters are surgical and introduce minimal artefacts.
Step 3: Noise Gate for Pauses
A noise gate silences audio below a threshold during pauses between speech.
Settings for spoken word: Threshold just above your noise floor, attack 1–5ms, hold 200–500ms, release 100–200ms. Set range to -80 to -90 dBFS rather than infinite attenuation, since a tiny amount of room tone sounds more natural than absolute silence.[2]
Gate limitations: A noise gate only fixes silence. It does nothing to noise present during speech. If background noise is audible behind your voice, you need spectral noise reduction.
Step 4: Spectral Noise Reduction (If Needed)
Spectral noise reduction analyses a sample of background noise and subtracts that spectral signature from the entire recording.
Capturing a noise profile: Find 2–3 seconds of pure silence, select it, use your tool's "Get Noise Profile" function, then apply to the entire file.
Conservative settings are essential:
- Reduction: 6–8 dB maximum per pass
- Passes: 1–2 maximum
- Sensitivity: Low to medium
Tools: Audacity's Noise Reduction (free, decent quality), iZotope RX Voice De-noise (industry standard, adaptive mode handles varying noise), Adobe Audition's Adaptive Noise Reduction.
What Does Over-Processing Sound Like?
Over-processing is worse than a slightly elevated noise floor. Listen for: musical noise / "water drops" (tonal artefacts from uneven removal), hollow or "underwater" sound (loss of natural ambience), metallic or robotic quality (harmonic distortion from removing too much spectral content). If you hear any of these, reduce the processing amount. ACX's human quality review rejects files for artefacts even if the technical specs pass.
What Is the Complete Noise Reduction Processing Order?
For maximum effectiveness with minimum artefacts:
- High-pass filter at 80 Hz
- Notch filters for tonal noise
- Spectral noise reduction (conservative)
- Noise gate for pauses
- Compression, then proceed to the normal mastering chain
- Gain adjustment to -20 dBFS RMS target
- Peak limiting below -3 dBFS
- Silence padding and MP3 encoding
- Verify final MP3 against all ACX specifications
After noise treatment but before loudness adjustment, measure your noise floor. It should be -65 dBFS or lower to survive the gain increase during loudness adjustment. For a quick reference of all eight ACX specs, see the ACX platform page. For common questions, see the FAQ.
ChapterPass manages noise floor automatically. Upload your chapters and get files that meet ACX's -60 dBFS requirement without manual processing.