To check and fix a batch of audio files for a client or platform loudness spec, start from the written spec: its integrated loudness target and tolerance, its true peak ceiling, and its format rules. Measure every file with a meter that follows ITU-R BS.1770, then correct the failures with a gain change to the target and, only where peaks would go over the ceiling, true-peak limiting, so the mix itself is not altered. Convert the sample rate and bit depth with proper dither, measure the delivered copies again, and send a report that shows each file passed. Free meters and command-line tools can do the measuring and normalizing; a batch tool adds the re-measurement and the report.
This guide explains what a delivery spec measures, the units, which spec applies to which job, how to measure and fix a batch, the extra checks for audiobooks, format rules and how to prove a delivery passed. Each section links to a deeper guide on this site. Spec values change, so always check the current version of your client’s document.
What is a delivery spec, and what does it measure?
A delivery spec is a written list of technical requirements a file must meet before a broadcaster, platform or client accepts it. Their QC desk checks it when files arrive, and files that fail come back.
Most specs cover the same groups:
- Loudness: an integrated loudness target with a tolerance, sometimes with short-term or momentary limits and a loudness range limit.
- Peaks: a true peak or sample peak ceiling.
- Format: sample rate, bit depth, channel layout, file type and, for compressed files, bit rate.
- Content rules: for audiobooks, noise floor and room tone; for ads, maximum length.
- Delivery details: file names and metadata.
Audio delivery QC checklist: 12 checks before sending turns these groups into a routine you can run on every job.
LUFS, LKFS, true peak and loudness range explained
| Term | What it is |
|---|---|
| LUFS | Loudness units relative to full scale. How loud audio sounds, measured with the K-weighted, gated method in ITU-R BS.1770 |
| LKFS | The same measurement under the name ATSC uses for US television |
| LU | A difference in loudness. 1 LU equals 1 dB, so tolerances are written as plus or minus 1 LU |
| Integrated loudness | The gated loudness of the whole file |
| Short-term and momentary | Loudness over a sliding 3-second and 400-millisecond window |
| True peak (dBTP) | The highest level of the reconstructed waveform, found by oversampling, which can be higher than the highest sample |
| Sample peak | The highest sample value in the file |
| Loudness range (LRA) | How much loudness varies across the file, per EBU Tech 3342 |
The difference between sample peak and true peak matters. A file can have every sample below 0 dBFS and still clip when it is converted to analogue or encoded to MP3, because the waveform between samples goes higher. That is why most modern specs set a true peak ceiling. LUFS, LKFS, dBTP and LRA explained goes deeper.
Which spec applies: ACX, EBU R128, ATSC A/85, podcasts or ads?
The spec comes from where the audio is going. The headline numbers for common targets, as stated in their own documents:
| Spec | Loudness target | Peak ceiling | Notes |
|---|---|---|---|
| EBU R 128 (broadcast) | -23 LUFS | -1 dBTP | European broadcast; see the tolerance note below |
| EBU R 128 s1 (short-form) | -23 LUFS | -1 dBTP | Adds a maximum short-term loudness for ads and promos |
| ATSC A/85 | -24 LKFS | -2 dBTP | US television |
| Apple Podcasts | About -16 LKFS, plus or minus 1 dB | -1 dBFS true peak | Recommended values for podcast audio |
| Spotify audio ads | -16 LUFS, plus or minus 1.5 | Not stated on Spotify’s ad spec page | Plus file type, bit rate and length rules |
| ACX audiobooks | RMS between -23 and -18 dB | -3 dB peak | Plus noise floor, room tone and MP3 rules |
Broadcasters often publish their own sheets on top of these, such as ARD, ZDF and ORF’s 2025 loudness delivery spec or UK DPP AS-11. Short-form content has its own rules in EBU R 128 s1, and Spotify’s audio ad specs vary by market.
One detail trips up many file deliveries: the tolerance. EBU R 128’s plus or minus 1 LU is written for cases where hitting the target is not practical, such as live programmes. A file-based delivery should land on the target itself. See EBU R 128 tolerance: 1 LU vs 0.2 LU explained.
How do I measure a whole batch?
Measuring one file in a meter plugin is easy. Measuring 40 chapters or 120 spots by hand is where mistakes happen.
Ways to batch-measure:
- ffmpeg. The free, open-source ffmpeg has filters that measure loudness to BS.1770, and the free ffmpeg-normalize tool wraps them for batches. Command-line, but reliable.
- Free loudness meters. Several free meters measure integrated loudness and true peak for files you load.
- DAW and editor tools. Many audio editors include loudness measurement and batch processing.
- A dedicated batch checker that marks each criterion of a named spec as pass or fail.
Whatever you use, make sure it measures true peak, not only sample peak, and that it follows the current BS.1770 method. For EBU R 128 work, see batch normalize audio to EBU R128 on Windows. For audiobooks, see the ACX batch checker for Windows.
How do I fix loudness without changing the mix?
The goal is to change the level, not the sound. In order:
- Measure each file’s integrated loudness.
- Apply a static gain change equal to the difference between the measured value and the target. This moves the whole file up or down and leaves the mix’s balance and dynamics alone.
- Check true peak after the gain change. Turning a file up raises every peak by the same amount.
- If true peak now exceeds the ceiling, apply true-peak limiting with a look-ahead limiter and oversampled detection, set to the ceiling. Good limiting touches only the few peaks that go over.
- Measure again. Limiting can lower integrated loudness slightly, so confirm the file still meets the target.
Avoid fixing loudness with heavy compression unless the client asks for it. Compression changes the mix. If a file needs more than a few decibels of limiting to reach the target, the master may need attention in the mix itself. How to fix true peak overs in WAV files covers the limiting step, and Audition Match Loudness alternative for Windows compares batch tools fairly.
Audiobook extras: noise floor, room tone and digital silence
ACX, the audiobook marketplace run by Audible, states its rules in older terms than LUFS:
| Requirement | ACX value |
|---|---|
| RMS level | Between -23 dB and -18 dB RMS |
| Peak level | No higher than -3 dB |
| Noise floor | No higher than -60 dB RMS |
| Room tone | No more than 5 seconds at the beginning and end of each file |
| File format | 192 kbps or higher constant bit rate MP3, 44.1 kHz |
| Channels | All mono or all stereo, consistently |
Two things catch narrators out. Digital silence (true zero) at the head or tail sounds like an edit error to QA; room tone, the quiet sound of the recording space, is expected instead. Noise floor is measured in the quietest part of the file, so a noisy room or a hum can fail a chapter that sounds fine in the middle. ACX’s human QA also listens, so passing the numbers is necessary, not a guarantee. Other audiobook distributors have their own lists; see Author’s Republic audio requirements.
Sample rate, bit depth, dither and file format
Format rules are simple but strict.
- Sample rate. Broadcast and post work usually want 48 kHz; music, audiobooks and many podcast hosts use 44.1 kHz. Convert with a high-quality resampler.
- Bit depth. Reducing bit depth, such as 24-bit to 16-bit, should use dither, typically TPDF, to avoid distortion on quiet passages. See dither 24-bit to 16-bit audio: when, which type and how.
- Container. WAV files have a 4 GB limit; long multichannel files may need RF64 or BW64. See WAV file over 4 GB? RF64 and BW64 explained.
- Metadata. Broadcast Wave files carry a bext chunk with description, origination and, in later versions, loudness values. Keep it when processing. See normalize BWF audio and keep bext metadata.
- MP3. Where a spec asks for constant bit rate, do not deliver variable bit rate.
How do I prove a delivery passed?
Clients and QC desks trust numbers they can see. A delivery report should include, for each delivered file:
- The spec it was checked against, by name and version.
- Before and after measurements: integrated loudness, true peak, loudness range and any other criteria.
- Pass or fail for each criterion.
- Format details: sample rate, bit depth, channels and container.
- A checksum, such as a SHA-256 digest, so anyone can confirm the file they received is the one you measured.
The most important habit is to measure the delivered files from disk, not the session or the files before export. Encoding and conversion can change the numbers.
How WavSpec fixes this
WavSpec loads a batch, checks each file against a named delivery spec, corrects what can be corrected without touching the mix, re-measures every delivered file from disk, and writes a report you can send with the files.
- What it checks. Integrated loudness, momentary and short-term maxima, loudness range, true peak and sample peak per ITU-R BS.1770-5 and EBU Tech 3341 and 3342; for audiobook and voice specs, RMS, noise floor and room tone at the head and tail; sample rate, bit depth, channel layout and container.
- What it corrects. Gain to the exact target, look-ahead true-peak limiting with 4x-oversampled detection, sample-rate conversion, bit-depth conversion with TPDF dither, float-to-PCM, MP3 delivery at a constant bit rate through the Windows encoder, and for audiobook specs, head and tail room tone, padded with the file’s own room tone, never synthesised noise.
- Built-in specs, each cited to its source and dated: EBU R 128 broadcast and s1 short-form, UK DPP AS-11, ARD/ZDF/ORF TV (2025), ATSC A/85 (2026), Apple Podcasts, Spotify audio ads, ACX, Author’s Republic and a generic 48 kHz / 24-bit template. You can also create and share your own.
- What you get. A report in PDF, HTML, CSV and JSON with before and after measurements, pass or fail per criterion, a loudness timeline per file and the SHA-256 digest of every delivered file. Broadcast Wave deliveries carry the verified loudness values in their bext metadata.
- Originals never overwritten. Corrected files go to a separate delivery folder. A watch folder and a command line are included for regular deliveries.
- Private. All processing runs locally and offline on Windows 11 (version 22H2 or later). No account, no cloud upload, no telemetry.
WavSpec has a free trial on the Microsoft Store, then a one-time purchase.
Get WavSpec from the Microsoft Store
Frequently asked questions
Are LUFS and LKFS the same?
Yes. They are the same BS.1770 measurement. EBU documents say LUFS and ATSC documents say LKFS.
What is the difference between true peak and sample peak?
Sample peak is the highest sample value. True peak estimates the highest level between samples by oversampling, and it can be higher.
Can I fix loudness without changing my mix?
Yes. Use a static gain change to the target, then true-peak limiting only where peaks would exceed the ceiling, and measure again.
What does ACX require for audiobooks?
RMS between -23 and -18 dB, peaks no higher than -3 dB, a noise floor no higher than -60 dB RMS, and 192 kbps or higher constant bit rate MP3 at 44.1 kHz.
More guides in this topic
- LUFS, LKFS, dBTP and LRA explained
- Audio delivery QC checklist: 12 checks before sending
- Batch normalize audio to EBU R128 on Windows
- EBU R 128 tolerance: 1 LU vs 0.2 LU explained
- EBU R 128 s1 short-form loudness explained
- ARD ZDF ORF loudness spec TPR 2025
- UK DPP AS-11 audio loudness: check before delivery
- Spotify audio ad specs: check a batch before upload
- ACX batch checker for Windows, with a report
- Author’s Republic audio requirements checklist
- How to fix true peak overs in WAV files
- Dither 24-bit to 16-bit audio: when, which type and how
- WAV file over 4 GB? RF64 and BW64 explained
- Normalize BWF audio and keep bext metadata
- Audition Match Loudness alternative for Windows
- All WavSpec topics