Every free ACX tool a narrator reaches for measures the file. None of the free ones fixes it. There are two families: the official ACX Audio Lab inside your ACX dashboard,[2] and the free third-party tools, which are the Audacity ACX Check plugin plus browser tools Tembrica (formerly Timbrica), Starsurfer Media, and OlaVoices.[4]
Here is the fact worth knowing up front. ACX Audio Lab does not measure noise floor. Its own help page says the tool "does not detect issues such as noise floor or editing errors."[1] Noise floor is one of the three level specs ACX rejects for, so the official tool has a blind spot exactly where narrators fail most often. The Audacity ACX Check plugin does report noise floor, along with RMS and peak,[4] but it reads sample peak, not true peak, and the reconstructed signal between samples can sit measurably higher.[5] ACX asks for peak levels less than -3 dB without saying which of the two readings it takes,[3] so a file that clears the plugin's sample-peak reading can still sit above -3 dB where it counts.
All of these tools tell you the numbers. None of the free ones changes the file on its own. That gap is what this article is about. The targets they are all reading against are the ones in the full ACX audio requirements: RMS between -23 and -18 dBFS, peak below -3 dBFS, and noise floor below -60 dBFS RMS.[3]
The ACX figures on this page were verified against the ACX help centre on 21 September 2026.
What does the ACX Audio Lab actually measure?
ACX Audio Lab is the official analyser, and it lives inside your ACX dashboard.[2] It reads the mechanical parts of the spec well: RMS, peak, bitrate, bitrate method, sample rate, mixed channels, duplicate files, and spacing between sections. For a fast sanity check on whether your loudness and format are broadly right before you submit, it does the job.
Then comes the gap. The tool does not measure noise floor, and it does not catch editing errors, extraneous noise, or dynamics problems.[1] That matters because noise floor is usually the first spec to flag for anyone recording outside a treated room. A file can sail through Audio Lab on loudness and format and still come back rejected on a hiss the tool never looked at. If that is where you are, fixing a noise floor above -60 dBFS is the guide to read next.
What does the Audacity ACX Check plugin measure?
The Audacity ACX Check plugin reports three numbers against ACX targets: RMS, peak, and noise floor.[4] That noise-floor reading is the reason many narrators reach for it over the official Audio Lab. It fills the gap Audio Lab leaves.
Two things are worth knowing. First, it is a separate download, a Nyquist analysis plugin, not something shipped with Audacity by default.[4] The Audacity ACX Check plugin settings guide walks through installing it and reading each result. Second, the plugin's noise-floor reading is the RMS level of the quietest 500 milliseconds in the selection.[4] That method can give a false pass on intermittent noise. A USB interface whine that sits above -60 dBFS during your speech but drops away in the one quiet passage the plugin samples will read as a clean floor. The number looks fine; the file is not.
Why can a file pass the ACX Check plugin and still get rejected?
This is the one that costs narrators days, so it is worth being precise. The plugin reads sample peak. Sample peak and true peak are not the same measurement, and ACX does not publish which one it reads.
Sample peak is the loudest single digital sample in the file. True peak estimates the actual analogue signal level between samples, using 4x oversampling to reconstruct the waveform that a listener's device will play back.[5] ITU-R BS.1770, the standard behind true-peak metering, documents the gap directly: sample-peak meters under-read the true peak, by a demonstrable 3 dB on an unfortunately phased tone at a quarter of the sampling rate, and "commonly several dBs" on real transients carrying significant high-frequency content.[5] ACX states only that peak levels must be less than -3 dB,[3] and its own help pages note that peak readings differ between DAWs, meters and plugins without naming the method it uses.[1] So a file that reads clean on the plugin's sample-peak meter can still be over the line where it matters. Target true peak under -3 dBFS and both readings pass.
If a file has cleared a free check and still come back rejected, a file that passes a free check but still gets rejected covers this and every other rejection reason in one place.
What do the free browser tools (Tembrica, Starsurfer, OlaVoices) measure?
Three browser tools show up most often, and all three measure the file in the browser with no upload. Tembrica reports RMS, peak, and noise floor per file, then compares chapters against each other and flags inconsistencies.[6] Starsurfer Media checks RMS, true peak, noise floor, sample rate, bitrate, and opening and closing silence in a single batch pass.[7] OlaVoices measures RMS, peak, and noise floor with no upload required, and is explicit that its peak reading comes from the decoded audio samples, the same ones a DAW reads, not a reconstructed true-peak estimate.[8]
All three are useful for a quick read, especially because your audio never leaves your machine. One caution applies to more than one of them. Tembrica links a "Fix in ACX Mastering" tool for chapters that fail its check,[6] Starsurfer separately sells a mastering service booked by call,[7] and OlaVoices advertises a paid audio cleanup service alongside its free measurement tool.[8] The quality of any of that paid mastering is the vendor's own claim, unverified here, so treat the free checks as ways to see the numbers, not as proof of what a paid step delivers.
What do none of the free tools measure or fix for you?
Line the gaps up honestly:
- Noise floor is missing from the official ACX Audio Lab entirely.[1]
- True peak is not measured by the Audacity plugin or by OlaVoices; both read sample peak instead,[4][8] which under-reads the level between samples,[5] so either can pass a file that is over -3 dB on a true-peak meter.
- Head and tail silence and cross-file consistency are patchy. The Audacity plugin does not check silence padding at all. Starsurfer checks opening and closing silence; Tembrica checks cross-chapter consistency; neither free tool covers everything the other does.[6][7]
- Not one free measuring tool masters the file itself. Where a fix exists, it is a separate paid product, not part of the free check.
That last point is the real line. Measuring is the easy half. You can gather every number from a stack of free tools and still be holding a file that fails, with no button anywhere in the free tier that repairs it. The specs behind all of this are in the full ACX audio requirements, and a noise floor above the line is fixed in the noise-floor guide.
Do these tools use the same analysis ACX uses to approve files?
No. The third-party tools approximate ACX's checks. The Audacity ACX Check plugin's own documentation is explicit about this: it is "intended only as an aid in achieving ACX acceptance," and even a straight pass "is NO guarantee of ACX acceptance."[4] They are not running ACX's own analyser, so their results can differ from what ACX reports when you actually submit. The true-peak gap above is the clearest example, but the principle is general: a green light from a free third-party tool is a strong hint, not a guarantee.
The official ACX Audio Lab is closer to the source, since it is ACX's own tool,[2] but it only reads part of the spec and skips noise floor.[1] So neither route gives you a complete, authoritative pass on its own. Read a free tool's result as good evidence, then make sure the specs it does not cover are handled too.
How do you go from a measurement to a fix?
A measurement tells you what is wrong. Fixing means changing the file so the numbers land inside the targets, and the order matters because each step moves the next.
The work is: adjust loudness into the RMS window, limit true peak under the ceiling, manage the noise floor below its line, convert to the required format and sample rate, and pad head and tail silence to spec. The catch is that these interact. Raising loudness to reach the RMS window raises the noise floor by the same amount, which is why a raw recording that measured clean can fail once it is brought up to level. That tension, and the way to work with it rather than against it, is the whole subject of fixing a noise floor above -60 dBFS. Get the order right and the file passes; get it wrong and fixing one number breaks another.
Which tool should you reach for, and when?
A plain guide:
- For a fast RMS, peak, and format sanity check inside your dashboard, open the ACX Audio Lab.[2] Just remember it will not tell you about noise floor.[1]
- When you want noise floor too, reach for the Audacity ACX Check plugin or a browser tool such as Tembrica, Starsurfer, or OlaVoices.[4] They read the three level specs, with the sample-peak caveat on the plugin and on OlaVoices.[5][8]
- When the numbers fail and you need the file repaired, not just graded, reach for a tool that masters as part of the free step, rather than a separate paid add-on. This is the step the free measuring tools do not do on their own. chapterpass measures all eight specs and masters the file to hit them, in your browser. You can run the free ACX check to see where your file stands before you commit to anything.
Common questions about the free ACX tools
Does the ACX Audio Lab check noise floor?
No. ACX Audio Lab reports RMS, peak, bitrate, sample rate and spacing, but its own help page states it does not detect noise floor or editing errors,[1] so measure noise floor separately before you submit.
My file passed the Audacity ACX Check but ACX rejected it. Why?
The plugin reads sample peak, not true peak.[4] ITU-R BS.1770 documents that sample-peak meters under-read the true signal level between samples, by a demonstrable 3 dB on an unfortunately phased high-frequency tone and "commonly several dBs" on real transients.[5] ACX asks for peaks under -3 dB and does not publish which reading it takes,[3] so a file that looks clean on the plugin can still be over the line. Target true peak under -3 dBFS and you satisfy either method.
Are Tembrica, Starsurfer and OlaVoices accurate?
They measure RMS, peak and noise floor in your browser and are useful for a quick read,[6] but they approximate ACX's checks rather than running ACX's own analyser, so treat a pass as a strong signal, not a guarantee.
Is there a free tool that fixes the problems it finds?
The free measuring tools stop at the number, or point you to a separate paid mastering step. chapterpass measures all eight ACX specs and masters the file to hit them, and your first file is free.
Can any tool catch clicks, mouth noise or pacing?
No automated tool catches those. ACX Audio Lab itself does not detect editing errors,[1] and they are human quality-review issues you fix by editing or re-recording, not by mastering.
Which free tool checks consistency across chapters?
Most free tools grade one file at a time. Tembrica is the one built to compare chapters directly and flag files that drift from the average,[6] though whether it repairs any inconsistency it finds is a separate, paid step.
Ready to submit?
Run the free ACX check to see exactly where your file stands, or let chapterpass master it to spec in one step.