You did everything right. Hours of recording, careful editing, a limiter on the master, clean exports. You uploaded to ACX and hit submit. Then the email came back: rejected.
The frustrating part is that ACX's rejection messages are often vague. "Audio quality does not meet requirements" could mean about six different things. This guide breaks down every technical reason submissions fail, explains exactly what ACX is actually measuring, and tells you how to fix each one before it costs you another round of edits.
What ACX Actually Checks
Every file you submit gets run through an automated QC process that measures four things. The values below come straight from the official ACX audio submission requirements:
| Parameter | ACX Requirement |
|---|---|
| RMS Level | -23 to -18 dB RMS |
| Peak | -3 dB maximum |
| Noise Floor | -60 dB RMS or lower |
| File Format | MP3, 192 kbps or higher, CBR, 44.1 kHz |
Four numbers. But the way they're measured is where most people get tripped up.
The 6 Reasons Your File Is Getting Rejected
Noise Floor Measured in the Wrong Place
This is the most common one, and it catches a lot of experienced narrators off guard. Most people check the noise floor at the very start of the file, see a nice clean -65 dBFS reading, and move on. Then ACX rejects it anyway.
Here's why: ACX measures noise floor in any silence segment throughout the entire file. That includes the half-second gap between sentences, the breath pause before a new paragraph, and the room tone at the end. If your air conditioning kicks in mid-session, or your mic self-noise creeps up over a long recording day, those mid-file silences will fail even if your intro sounds perfect.
Confusing True Peak with Peak
Your DAW's standard peak meter shows one number. ACX requires peaks at -3 dB or below, and the safe way to verify that is with a True Peak meter (dBTP). Sample peak and True Peak are not the same measurement, and the difference matters.
True Peak catches inter-sample peaks, which are clipping events that happen during digital-to-analog conversion. A file that shows -3.2 dB on your regular peak meter can easily exceed -3 dBTP once it's been processed by ACX's system. You're submitting something you think is clean, and it's failing a check you didn't know existed.
Measuring Loudness Across the Whole Book Instead of Per Chapter
When you submit an audiobook, ACX checks each chapter file independently. If you measure your level across a full batch export of all chapters combined, you might hit -20 dB RMS overall while individual chapters vary anywhere between -17 and -24 dB RMS.
A quiet prologue, a slow chapter without much action, or a session recorded on a different day with slightly different gain settings: any of these can push an individual chapter out of range even when the average looks fine.
Measuring LUFS Instead of RMS
Modern loudness workflows revolve around LUFS, based on the ITU-R BS.1770 standard, so a lot of narrators verify their chapters with a LUFS meter. But ACX specifies its target in RMS: -23 to -18 dB RMS. The numbers often look similar for spoken word, but they're calculated differently: LUFS applies frequency weighting and gating, while RMS is a straight signal power average.
If you're targeting the range with a LUFS meter, you can sit at -19 LUFS and still fall outside the RMS window - passing or failing by a margin that's invisible in your workflow.
Edit Points Causing Micro-Distortion
This one shows up less often, but it's particularly sneaky because it sounds completely fine in your DAW. Hard cuts between takes can create tiny clicks or transient spikes when the waveform isn't at a zero crossing at the edit point. ACX's automated QC sometimes flags these as distortion.
Heavily edited chapters with a lot of punch-ins and comped takes are most vulnerable to this.
Wrong File Format or Export Settings
ACX accepts exactly one upload format: MP3 at 192 kbps or higher with CBR (Constant Bit Rate), at 44.1 kHz. Record and edit in WAV all you want, but the uploaded file must be MP3. Submitting variable bit rate MP3 or 48 kHz audio fails the technical check regardless of how good your levels are.
A Pre-Submission Checklist Worth Bookmarking
- Measure RMS on each chapter individually (-23 to -18 dB RMS)
- Check True Peak with a meter that explicitly says "True Peak"
- Measure noise floor mid-file in a silence between sentences, not just the start
- Confirm export format: MP3, 192 kbps or higher, CBR, 44.1 kHz
- Check for 0.5 to 1 second of silence at the head and 1 to 5 seconds at the tail
Why This Gets Hard at Scale
For a single short chapter, running these checks manually is fine. For a 20-chapter audiobook, it takes a while, and the more you do manually, the more likely a missed step or a misread number becomes. Checking chapter 18 after four hours of editing is not the same as checking chapter 1 fresh.
The other issue is that ACX's automated system doesn't always behave exactly the way things look in your DAW. Subtle inter-sample peaks, noise sitting right on the -60 dB RMS boundary, or a single outlier chapter in a long book can mean a full rejection after days of work.
Running an automated check against the exact ACX spec before you upload removes that uncertainty. If it says pass, you submit with confidence. If it finds a failure, you know exactly which chapter and which parameter to fix before ACX ever sees the file.