Standards & Delivery
Make media that holds up in the real world.
Primary-source references, transparent methods, and practical tools for measuring, checking, and delivering accessible media.
Audio delivery
Separate broadcast targets, podcast working points, and music normalization references.
Accessibility QA
Plan captions, readable text, audio description, keyboard access, and human review.
Provenance & AI
Understand content credentials, synthetic-media disclosure, and platform policy changes.
Production reference
Browse practical tools and guides by the media job you need to complete.
How to read this library
Claims are labeled by confidence
Standards and platform policies change. Each technical tool should state its source, scope, review date, and method. Until a calculation has public fixtures and review evidence, it should not be described as formal compliance certification.
Reference
Summarizes a primary standard, policy, or published platform recommendation.
Estimated
Useful for planning or education, but not a substitute for a formal delivery meter or human review.
Validated
Backed by published fixtures, tolerances, and a documented comparison or review process.
Experimental
A working prototype whose behavior or method is still being evaluated.
Primary references
Read on 21 September 2026. Each entry says what the document establishes. The figures on this page come from these documents and not from a summary of them.
- ITU-R BS.1770-5, titled Algorithms to measure audio programme loudness and true-peak audio level. The ITU approved this version on 22 November 2023 and lists it as in force. The loudness targets below refer to this measurement.
- EBU R 128, version 5, November 2023. It sets the programme loudness target at -23.0 LUFS. It permits a tolerance of ±1.0 LU where the target is not practically attainable, such as live programmes, and allows ±0.2 LU in quality control to account for measurement error. It says the true peak of a programme shall not exceed -1 dBTP during production.
- Apple Podcasts audio requirements. Apple recommends an overall loudness of around -16 dB LKFS with a tolerance of ±1 dB, and a true peak no higher than -1 dB FS, calculated according to ITU-R BS.1770-5. RSS feeds accept MP3 or AAC. The page words the loudness figures as a recommendation.
- WCAG 2.2, a W3C Recommendation dated 12 December 2024 on the page. It defines three conformance levels, A, AA and AAA, and adds nine success criteria to WCAG 2.1. This site targets level AA.
- C2PA specifications. The index lists version 2.4 as current. C2PA describes its work as technical standards for certifying the source and history, or provenance, of media content.
Review and corrections
Every standards-sensitive page should show an owner, reviewer, sources, applicable version, last reviewed date, supported scope, and known limits. Report a factual or accessibility issue through the corrections process. Material corrections should be documented in the page changelog.
How to use this library
Delivery specifications come from broadcasters, platforms and standards bodies who each solved a slightly different problem. A loudness target for broadcast television and a loudness target for a podcast are both correct, and they are not the same number. Applying one where the other belongs is the most common reason a technically clean file comes back rejected.
The pages here separate those contexts before giving you a number. Where a value comes from a published specification, the page cites the specification. Where a value reflects common practice rather than a written rule, the page says so plainly.
Four confidence labels run across the site. Reference means the value comes from a published standard. Estimated means we derived a figure and show the method. Validated means we tested against real files. Experimental means the area is still moving and the guidance will change.
Specifications get revised, and platforms change requirements without announcing them. Check the review date on any page carrying a number you plan to deliver against, and confirm against the spec you were actually given. Compatibility with a standard described here is not endorsement by the organisation who wrote it.
Two correct numbers for one measurement
A broadcast programme and a podcast episode are measured the same way and delivered at different levels. Both documents use programme loudness, the loudness averaged over the whole programme, and both point to ITU-R BS.1770 for the method. EBU R 128 sets the target at -23.0 LUFS. Apple recommends around -16 dB LKFS. LUFS and LKFS are one measure under two names. The EBU says so in a footnote to R 128.
The two targets sit 7 loudness units apart. A loudness unit, written LU, is the step used for differences in loudness. A mix finished at -23.0 LUFS for broadcast measures 7 LU below Apple's recommended level. A podcast finished at -16 LKFS measures 7 LU above the EBU target. Neither file is wrong. Each fits the system it was made for.
The peak limits are closer. R 128 says the true peak of a programme shall not exceed -1 dBTP during production. Apple says its true-peak value should not exceed -1 dB FS. True peak is the highest value the signal waveform reaches, including between the samples of a digital file. Apple gives clipping during encoding as its reason for the limit. The EBU describes the true-peak measure as the check against the upper technical limit of the signal chain.
To measure a file against either target, use the LUFS Checker. For the working points by format, use the podcast loudness cheat sheet. Both label their figures with the scope labels above.
How this library treats its own numbers
Where a page cites a source, it links the primary document and says what the document establishes. Where no primary record could be reached, the Sources section says so and names the record that would settle the claim. It does not cite a summary in its place.
A modified date on a page says when the page last changed. It does not claim a person reviewed the page on that day. A check date appears only where every figure on the page was verified against the source.
Price figures are approximate and in US dollars. A page that quotes a price band keeps the band round, so it does not suggest a precision nobody checked.
None of the tools here certifies compliance with a broadcaster or platform specification. Confirm against the specification you were given before you deliver. The tools run in your browser. Files you open in them stay on your device and are not uploaded to us.
Vocabulary
Programme loudness. The integrated loudness over the whole duration of a programme, given in LUFS. It is one number for the whole item, not a reading at one moment. Broadcast and podcast targets both use it.
LUFS and LKFS. Loudness Units, referenced to Full Scale. LKFS is the name ITU-R BS.1770 uses and LUFS is the name the EBU uses. They describe the same measure. A lower number means a quieter programme.
LU. Loudness Unit. The step used to describe a difference or a tolerance in loudness. A target of -23.0 LUFS with a tolerance of ±1.0 LU accepts readings from -24.0 to -22.0 LUFS.
True peak. The maximum value of the audio waveform in continuous time, which includes points between the digital samples. Measured in dBTP. A limit on it keeps the signal from clipping when the file is converted or encoded.
Provenance. The recorded source and history of a piece of media, such as who made it and what edited it. C2PA defines a standard way to attach that record to a file.