ACX Specs, Explained in Plain Language
ACX Specs, Explained in Plain Language
If you’re recording an audiobook for Audible, you’ll hear the term “ACX specs” thrown around a lot. Here’s what they actually mean, why they exist, and what happens if your file doesn’t hit them.
What ACX Is
ACX (Audiobook Creation Exchange) is Audible’s platform for producing and distributing audiobooks. Every file uploaded to ACX gets run through an automated quality check before it’s accepted. If your file falls outside spec, it gets kicked back. No exceptions, no manual review on the first pass.
The Four Numbers That Matter
RMS levels: -23dB to -18dB. RMS is your average loudness across the file, not your peak. This range keeps your narration loud enough to hear comfortably without being so hot it distorts. Too quiet and listeners have to crank their volume; too loud and you risk clipping. If you fail it: you re-run the file through a limiter or gain stage and resubmit. Usually a same-day fix. Why it doesn’t happen here: tracking level and RMS aren’t the same measurement. I track with peaks around -12dBFS to -8dBFS, well below 0dBFS, so there’s headroom before anything clips. RMS is a separate average-loudness measurement, set later at mastering with a limiter to land the file in the -23 to -18dB range. The peak headroom left during tracking is what makes hitting that RMS target possible without distortion.
Peak level: under -3.0dBTP. This is your loudest single moment, measured as true peak (accounts for inter-sample peaks that can occur during MP3 encoding, not just the raw waveform). Stay under -3.0dBTP and you have headroom so the file doesn’t clip when it’s converted to the final format. If you fail it: same as RMS, this is a mastering-stage fix. Pull the peak down, re-export, resubmit. Why it doesn’t happen here: the mastering chain ends on two limiters set to a -3.5dBTP ceiling specifically to leave margin below ACX’s -3.0dBTP cutoff, so there’s no borderline pass.
Noise floor: below -60dBRMS. This is how quiet your silence actually is. Room tone, HVAC hum, electrical noise from your interface, mouth clicks between sentences. All of it gets measured. A noisy room with great performance still fails here. If you fail it: this is the one that hurts. You’re not adjusting a slider, you’re finding the noise source in the room and treating it, which sometimes means re-recording sections entirely. Why it doesn’t happen here: the room was built to control noise before a single session ever ran in it. Sessions average -89dBRMS, nearly 30dB under the -60dB minimum, so there’s no ambiguity at the measurement stage. Worth knowing: ACX Audio Lab doesn’t actually report this number, it’s pass/fail only, so getting the real figure means running the file through a third-party meter.
File format: 192kbps or higher CBR MP3, 44.1kHz. ACX’s floor is 192kbps, constant bitrate, not variable. Higher bitrates are allowed but don’t add anything a listener can hear. Sample rate is 44.1kHz. None of this is a creative choice, it’s a delivery requirement. If you fail it: re-export with the correct settings. This one’s rarely a real failure, usually just someone exporting VBR instead of CBR, or forgetting to change the default bitrate. Why it doesn’t happen here: every audiobook file is delivered at 320kbps CBR, 44.1kHz, as a fixed part of the mastering workflow. It’s not a setting I have to remember to check.
This is the tradeoff nobody tells you about upfront: ACX specs are pass/fail on the back end, but the work to hit them happens on the front end, at the mic, in the room, before mastering ever starts. Three of these four specs get decided the moment you hit record, not when you export the final file.
What This Costs to Get Right
Pricing this out before you record saves a lot of guesswork. I built a cost calculator that takes your manuscript’s word count and returns a real range: it works off 7,800 to 9,000 words per finished audio hour, then applies a 1.5x to 2x studio-time multiplier on top of that to account for retakes, direction, and pacing. Recording runs $50 per studio hour, mastering runs $150 per finished hour. A 94,000-word manuscript, for example, lands between roughly $2,375 and $3,000 start to finish. Run your own word count through it below.
What I Actually Check
I run every file through ACX Audio Lab before it goes out the door, and that result is the final word, not my ear, not a plugin readout. I’ve seen files that sound clean to the ear fail on noise floor because of something under -60dB that isn’t audible in normal listening but shows up in the measurement. The spec doesn’t care how it sounds to you. It cares what the numbers say.
If you’re recording your own audiobook and self-mastering for ACX submission, know the room you’re recording in matters as much as the mic you’re using. You can EQ and compress your way to good RMS and peak numbers after the fact. You can’t EQ your way out of a bad noise floor without also degrading the voice underneath it.