ACX Noise Floor (−60 dB) & Audiobook Cleanup Guide: Pass QA Without Gating
Every official ACX Audio Submission Requirement explained in plain English, why raising loudness pushes home-studio noise floors past −60 dB, why hard noise gates and generative speech enhancers trigger rejections, and a 3-step workflow that keeps 100% human narration intact.
Updated November 15, 2026Michael NguyenTested November 2026
Try LUFS meter from Audo
More options
Quick answer
Audible's ACX Audio Submission Requirements state that every audiobook chapter must measure between −23 dB and −18 dB RMS, peak no higher than −3.0 dB, and keep a maximum −60 dB RMS noise floor — three numbers you can check 100% free on Audo AI's LUFS meter and meet without hard gating using Audo AI's non-generative Remove noise and Boost volume (Audiobook target). On a 10–100% slider, Audo AI's Remove noise cleans 1 minute of narration in ~7s, 20 minutes in ~33s, and a 1-hour chapter in under 2 minutes for $0.05 a minute ($3.00 an hour list, or $1.88 to $2.40 an hour on non-expiring packs from $5, with 3 downloadable 1-minute free previews a day and no sign-up), while Boost volume masters chapters to ACX specs for $0.01 a minute ($0.60 an hour).
| ACX requirement | Official ACX target | Common home-booth failure | How to meet & verify on Audo AI |
|---|---|---|---|
| Integrated chapter loudness (RMS) | Between −23 dB and −18 dB RMS per file | Quiet raw recordings (−26 dB RMS) fail low; podcast normalizers (−16 LUFS) overshoot −18 dB RMS | Switch Normalize volume OFF in Remove noise, then run Boost volume with Target set to Audiobook ($0.01/min; first 3 min free) |
| Maximum peak ceiling | No higher than −3.0 dB | Laughs, shouted dialogue, or un-limited plosives hit −1 dB or 0 dB | Boost volume's Audiobook target limits peaks at or below −3.0 dB; verify Peak and 0 Clipped samples free on LUFS meter |
| Maximum noise floor & natural room tone | No higher than −60 dB RMS, with 0.5–1s head and 1–5s tail room tone | Adding +6 dB gain lifts a −63 dB room floor to −57 dB; hard noise gates create −∞ dB dead dropouts | Run non-generative Remove noise at 75%–85% on the 10–100% slider to drop a −52 dB floor below −65 dB while keeping natural room tone |
| Delivery format, length & channels | 44.1 kHz, 192 kbps (or higher) CBR MP3; under 120 min/file; all mono or all stereo | Variable bit rate (VBR) exports or mixing mono and stereo chapters across one book | Export 192 kbps 44.1 kHz MP3 on the final pass and verify sample rate, channels, Average level, Peak, and Noise floor free on File info or LUFS meter (20 files/day) |
Hear the difference
Play Original, then Cleaned.
Cleaned removes the chatter and dishes.
Made with Remove noise
Official ACX Audio Submission Requirements in plain numbers
Unlike Spotify or Apple Podcasts (which measure loudness in LUFS), Audible's Audiobook Creation Exchange (ACX) grades every uploaded chapter, opening credit, closing credit, and 1-to-5-minute retail sample against three strict root-mean-square (RMS) and peak thresholds. Every file must measure between −23 dB and −18 dB RMS across the whole chapter, keep its loudest single Peak no higher than −3.0 dB, and hold its quietest pauses to a Noise floor no higher than −60 dB RMS.
Formatting rules sit right beside the decibel window: each uploaded file must hold a single chapter or section under 120 minutes with the chapter header read aloud, start with 0.5 to 1 second of clean room tone at the head and finish with 1 to 5 seconds of room tone at the tail (never exceeding 5 seconds), and export at 44.1 kHz as a 192 kbps or higher Constant Bit Rate (CBR) MP3 — never Variable Bit Rate (VBR). Across the entire book, every chapter must stay consistently all mono or all stereo.
Why home booths fail ACX QA: 'noise floor too high' vs. 'noise floor too low' & AI resynthesis
First-time narrators recording in a walk-in closet or spare bedroom almost always run into one of two opposite ACX rejections. The first is a 'noise floor too high' rejection between −52 dB and −58 dB RMS (above ACX's -60 dB limit) after mastering. Gain moves your voice and your room hiss together: if you record cautiously at −26 dB RMS with a seemingly safe −62 dB room floor from an HVAC vent, laptop fan, or USB preamp, adding +6 dB of makeup gain to hit −20 dB RMS lifts that noise floor straight to −56 dB RMS — an instant ACX QA failure.
The second trap happens when narrators over-correct with hard noise gates, Audacity's Silence Audio command, or generative voice enhancers. Clamping pauses to dead digital silence (−∞ dB) triggers 'noise floor too low' warnings and audible gating pump where room tone vanishes abruptly between words, violating ACX's requirement for natural room tone. Worse, running chapters through generative speech resynthesizers — such as Adobe Podcast Enhance or Descript Studio Sound — rebuilds vocal formants from an AI voice model rather than filtering noise out of your recording, smoothing away acting micro-inflections and risking rejection on platforms that require authentic human narration.
Failure 1: Noise floor climbs past −60 dB after RMS gain
Boosting a quiet chapter into the −23 to −18 dB RMS window raises PC fan whine, fridge rumble, and mic self-noise by the exact same decibel amount.
Failure 2: Hard noise gates create unnatural −∞ dB dropouts
Zeroing out pauses between sentences sounds like a broken telephone line next to live speech and violates ACX head/tail room-tone rules.
Failure 3: Generative voice enhancers alter human vocal cords
Speech resynthesis replaces your real timbre with synthetic formants; human-narrated audiobooks need non-generative noise filtering instead.
The 3-step Audo AI workflow to pass ACX QA with 100% real human narration
To drop your closet or booth noise floor safely below −60 dB RMS without gating dropouts or synthetic voice alterations, run your raw lossless WAV chapters (up to 10 GB and 10 hours) through three non-destructive steps on Audo AI:
Step 1: Run Audo AI's Remove noise at 75%–85% with Normalize volume OFF
Upload your raw chapter WAV to Remove noise (or Audio denoiser), set the 10–100% Noise reduction slider to 75% to 85%, and switch Normalize volume OFF on the main panel. Leaving Normalize volume on targets −16 LUFS podcast loudness, which overshoots ACX's −18 dB RMS ceiling; turning it off while setting 75%–85% drops a −52 dB noise floor past −65 dB while preserving subtle natural room tone and 100% of your real human vocal performance (~7s for 1 min, ~33s for 20 min, under 2 minutes for 1 hour; video rehearsals finish in ~10s for 1 min, under a minute for 10 min, ~2 minutes for 20 min, and under 4 minutes for 1 hour).
Step 2: Master chapter level in Boost volume set to Audiobook ($0.01/min)
Drop the cleaned WAV into Audo AI's Boost volume and set Target to Audiobook ($0.01 a minute = $0.60 an hour list, or $0.38–$0.48 an hour on packs from $5; first 3 minutes free). Audiobook places your chapter squarely inside ACX's −23 to −18 dB RMS window, caps peaks at −3.0 dB, reports whether all ACX rules pass, and lets you export your final 192 kbps 44.1 kHz MP3 under More options.
Step 3: Verify every chapter 100% free on LUFS meter or File info
Before uploading to ACX, drop each finished chapter into Audo AI's LUFS meter or File info (always 100% free for up to 20 whole files a day with no sign-up) to confirm Average level (−23 to −18 dB RMS), Peak (−3.0 dB or lower), Noise floor (below −60 dB), and 0 Clipped samples.
When Audacity's manual chain or iZotope RX fits better vs. Audo AI
Choose the mastering setup that matches your booth acoustics and schedule:
Audacity + Filter Curve, Loudness Normalization & Limiter (free)
When your treated vocal booth already sits below −65 dB RMS after gain and you simply want an offline desktop macro with zero per-minute cost.
iZotope RX Standard ($399 list)
When you edit audiobooks full-time and also need Mouth De-click and Spectral Repair to scrub wet saliva clicks and stomach growls alongside room hiss.
Audo AI Remove noise + Boost volume + free LUFS meter
When HVAC, computer fans, or preamp hiss push your chapters to −52 to −58 dB RMS, you have no clean noise-profile pause, and you want fast non-generative cleanup plus Audiobook RMS mastering for ~$3.60 per finished hour list ($0.05/min + $0.01/min, or ~$2.26–$2.88/hr on non-expiring packs from $5) with free whole-chapter ACX verification.
Clean audiobook noise floors and verify ACX specs free
Lower home-studio hiss below −60 dB RMS without gating or voice resynthesis in Audo AI's Remove noise (10–100% slider, first minute free), then verify RMS, Peak, and Noise floor 100% free in the LUFS meter.
Questions
What are the official ACX audio requirements for noise floor, RMS, and peak?
ACX requires every audiobook chapter to measure between −23 dB and −18 dB RMS, peak no higher than −3.0 dB, and maintain a noise floor no higher than −60 dB RMS as a 44.1 kHz, 192 kbps CBR MP3 — all checkable 100% free for up to 20 whole files a day on Audo AI's LUFS meter or File info. Each file must also stay under 120 minutes, keep consistent mono or stereo channels across the book, and include 0.5–1s head and 1–5s tail room tone.
How can I run a free ACX check online without installing Audacity plugins?
Drop any WAV or MP3 chapter up to 10 GB into Audo AI's LUFS meter or File info in your browser to check Average level (RMS), Peak, Noise floor, and Clipped samples 100% free (up to 20 whole files a day, no sign-up). If your noise floor fails above −60 dB, clean it with Audo AI's Remove noise and re-level with Boost volume set to Audiobook.
Why does ACX flag 'noise floor too low' or reject AI-enhanced audiobook audio?
Hard noise gates clamp pauses to dead digital silence (−∞ dB) that ACX flags against natural room tone, while generative enhancers like Adobe Podcast Enhance resynthesize human vocal formants; using Audo AI's non-generative Remove noise at 75%–85% on its 10–100% slider avoids both issues. It filters out fan and preamp hiss while keeping genuine room tone and 100% authentic human narration.
Why should I turn off Normalize volume in Remove noise before ACX mastering?
On Audo AI, Remove noise's Normalize volume toggle on the main panel defaults to −16 LUFS for podcasts, which sits louder than ACX's −18 dB RMS ceiling. Switch Normalize volume OFF when running Remove noise, then run Audo AI's Boost volume with Target set to Audiobook so your chapter lands cleanly inside −23 to −18 dB RMS with a −3.0 dB peak ceiling.
How much does it cost to clean and master a 1-hour audiobook on Audo AI?
Cleaning a 1-hour recording in Audo AI's Remove noise takes under 2 minutes and costs $3.00 list ($0.05/min, or $1.88 to $2.40 on non-expiring packs from $5), plus $0.60 ($0.01/min) to set ACX loudness in Boost volume, while whole-file checks on Audo AI's LUFS meter are 100% free with no subscription.