Average level (RMS)
−23 to −18 dBFS
−600 dBFS
How loud the chapter is overall. Almost every rejection starts here.
ACX chapter check
Check and fix your audiobook chapter
This measures the numbers ACX checks, tells you which fail and by how much, then fixes them and exports a compliant MP3. The tools on this site never upload your audio: every check and fix runs in this tab.
- Checks
- Fixes
- Nothing uploaded
- Built for narrators
Check a chapter
No file to hand? Try a sample chapter:
Seven seconds each, synthesized by this site rather than recorded, so nobody’s voice is in them.
Ready.
- Fixes it, not just checks it
- Room tone checked
- ACX · Spotify · INaudio · Author’s Republic
- Whole-book batch
- Nothing uploaded
Why narrators trust the result
- Measured, then fixedLevel, peaks and noise floor corrected in the order the numbers allow.
- Re-measured before you get itThe MP3 it wrote is decoded and checked again. No pass, no download.
- Honest about re-recordsWhen a noise floor cannot be fixed without damage, it says so.
- Four platforms’ published numbersACX, Voices by INaudio, Spotify for Authors and Author’s Republic.
From the booth to the upload
One studio, three stages
-
01 · Record
Before the red light
Test the room and the microphone, and know the length of the job before you read a word.
-
02 · Check and fix
Every chapter, to the spec
Measure the numbers ACX rejects for, then correct them, the room tone and the noise.
-
03 · Deliver
The whole book, consistent
Put accepted chapters in order, resolve pickups, check credits and the sample, then export a complete package.
Everything here, in the order you would use it
Twelve tools, one browser tab
Each one does a different job and runs entirely on this device. Your recordings stay on your device from the first check to the finished download.
- Before you record Room test Ten seconds with your microphone tells you whether the space can hold a book at all. Gives a verdict on your room
- Before you record Microphone test Which microphone the browser hears, at what rate, and whether your level is right. Gives level, peak, noise and a playback
- Before you record Book calculator Open the EPUB, PDF, DOCX or TXT itself and see what the job actually is. Gives finished hours, a budget and a narration script
- Per chapter · you are here Chapter checker and fixer The numbers measured against ACX, Spotify, INaudio or Author’s Republic, then corrected. Gives a compliant 192 kbps MP3, re-measured before you get it
- Per chapter Room tone checker Seconds of room at the head and tail against every platform, then topped up or trimmed. Gives an MP3 with both ends in range, re-measured
- Per chapter Remove background noise Learns the hum or hiss from your own quiet passage and subtracts it from the whole file. Gives a cleaner MP3 and the before and after
- Per chapter Trim audio Type a start and end in seconds and keep only that part, at the file’s own rate. Gives the trimmed MP3 beside the original
- The whole book Book workspace Order chapters, review pickups and accept each recording before preparing the complete delivery. Gives a saved project, pickup reports and a delivery package
- The whole book Batch checker Every chapter against ACX and against the book’s own median, so one loud chapter cannot hide. Gives one ZIP of fixed chapters, a CSV and a JSON report
- The whole book Audio joiner Parts in the order you choose, mismatched rates and channels reconciled first. Gives one MP3, or an M4B with a chapter per part
- A different job Podcast loudness LUFS rather than RMS: the ruler podcasts and streaming use, with each platform’s own target. Gives a normalized MP3, measured again after encoding
- A different job Audio converter WAV, AIFF, MP3 or M4A in; MP3 or uncompressed WAV out, keeping the source rate. Gives the converted file and what the encode cost
Also here: three checking aids that read your manuscript back to you, which start with a Google sign-in, and a developer API for repeatable production workflows. Compare Starter and Pro on the developer plans page.
- Hear your manuscriptYour text read back, to catch what the eye skips.Google sign-in required
- Pronunciation guideThe names and hard words, and how each one is said.Google sign-in required
- Proof-listenWhere the recording and the manuscript part company.Google sign-in required
- Developer APIThe same check and fix over HTTP, for your own pipeline.Starter and Pro plans
The five it rejects for
What it checks, before you give it anything
Each scale below shows the whole range the number can land in, with the band ACX accepts marked on it. The buttons play clips this page synthesizes, not recordings, each scaled so its measured level is exactly the figure on the button.
Peak
−3 dBFS or lower
−600 dBFS
The loudest single moment. Usually a plosive, and it is what makes the level fix non-trivial.
Noise floor
−60 dBFS or lower
−900 dBFS
The room between your words. This is the one that is sometimes not fixable, and we say so.
Room tone at both ends
1 to 5 seconds
0 s10 s
And when it is missing we build a second of it from your own quietest half second, not from digital silence.
Encoding
44.1 kHz · 192 kbps · CBR · MP3
Read from the file's own frame headers, every frame, because “constant” is a claim about all of them.
ACX audio requirements at a glance, and what is fixed here
| Requirement | ACX asks for | Fixed here | How |
|---|---|---|---|
| Average level (RMS) | −23 to −18 dB RMS | Yes | Set to about −20.5 dB, the middle of the window. |
| Peak | −3 dB or lower | Yes | Limited to −3.5 dB, half a decibel under the ceiling. |
| Noise floor | −60 dB RMS or lower | Sometimes | The gaps between words are lowered, and when that is not enough, noise under the voice is subtracted too, as long as speech is at least 18 dB above the room. Below that, the answer is a re-record. |
| Room tone | No more than 5 s at each end; 1 to 5 s recommended | Yes | Topped up from your own quietest half second, or trimmed. |
| File format | MP3, 192 kbps or higher, constant bit rate, 44.1 kHz | Yes | Written as a 192 kbps constant-bitrate MP3 at 44.1 kHz, then decoded and measured again. |
| Running time | 120 minutes or less per file | No | Splitting a chapter is an editorial decision, so it is left to you. |
Source: the ACX audio submission requirements. No meter measures pronunciation, extraneous sounds or consistency; ACX’s human Quality Control judges those.
The trap
Why raising the volume is not the fix
Gain moves all three numbers by the same amount. That is why a file comes back rejected for a different reason than last time.
Above a dashed line is a rejection. Only the level lands where it should; the peak clips and the floor rises with it.
So it corrects in the order the numbers allow
A chapter recorded at −27 dB needs about six decibels. This corrects the floor first — against the level it has to reach before the gain — and then limits the peaks on the signal that actually gets encoded.
Then it re-opens the MP3 it just wrote, decodes it and measures it again. If the file it produced does not pass, it says so and offers no download.
Straight answers
Questions narrators actually ask
What are the ACX audio requirements?
ACX requires RMS between −23 and −18 dB, peaks no higher than −3 dB, a noise floor of −60 dB RMS or lower, no more than five seconds of room tone at each end, and a 192 kbps or higher constant-bitrate MP3 at 44.1 kHz, with each file 120 minutes or shorter. It recommends one to five seconds of room tone. This page measures every one of them from your file's own bytes.
Does it fix the file, or only check it?
It fixes it. When a number fails and can be corrected, one button corrects what failed — the level, the peaks, the noise floor, the room tone or the format — writes a 192 kbps constant-bitrate MP3 at 44.1 kHz and measures that MP3 again. If the corrected file does not pass, there is no download. When it cannot fix something, such as a noise floor too close to your voice or a chapter over 120 minutes, it says so instead of offering a file.
Is my audio uploaded when I check it?
No. The tools on this site never upload your audio: decoding, measurement and the fix all run inside your browser tab. That is why a manuscript under NDA can be checked here at all. The separate, opt-in developer API processes only the files its key holders send to it.
If every check passes, will ACX accept my audiobook?
Not necessarily. Automated analysis proves the measurable specs; ACX's human Quality Control also judges pronunciation, extraneous sounds and consistency that no meter measures, and ACX's own Audio Lab FAQ says passing analysis does not guarantee approval.
Why does my DAW show a different RMS than ACX?
ACX's own FAQ notes that RMS can be measured in multiple ways, so DAW meters legitimately disagree with each other. This tool measures whole-file RMS in dBFS the way the requirements describe, then re-measures its own encoded output the same way.
Sources: the ACX audio submission requirements and the ACX Audio Lab FAQ.