Technique · media and documents · WCAG 1.2 and EN 301 549 clause 10
Video, audio and PDF — three places where accessibility leaks
A site can pass a code audit and still be unusable, because the content that matters sits in a video with no captions or in terms and conditions saved as a scan. Media and documents have criteria of their own: WCAG 1.2 for media, clause 10 of EN 301 549 for files that are not web pages. Both can be checked in under an hour, and both cost more to fix afterwards than to do right the first time.
Media: what WCAG actually requires
| Criterion | Level | Applies to | What you must provide |
|---|---|---|---|
| 1.2.1 | A | audio only (a podcast) or video with no sound | a text transcript; for silent video, a text description or an audio track |
| 1.2.2 | A | recorded video with sound | captions — synchronised, corrected, with speakers identified |
| 1.2.3 | A | recorded video | audio description or a full text alternative |
| 1.2.4 | AA | live streams | live captions (respeaking or stenography; automation alone rarely suffices) |
| 1.2.5 | AA | recorded video | audio description — here a text alternative no longer discharges the duty |
| 1.4.2 | A | sound that starts by itself and plays for more than 3 s | a way to pause or mute it independently of the system volume |
Level AA — the one the standard requires — means in practice: captions on everything with sound, transcripts for podcasts, audio description for video whose picture carries meaning. A film in which the narrator says everything you can see needs no audio description; a tutorial saying “click here” does.
Captions, transcript, audio description — not synonyms
Captions
Synchronised text of the speech and of meaningful sound (“[glass breaking]”), with speakers identified. For deaf users — and for everyone watching with the sound off, which is most mobile traffic.
Burned-in vs. separate
Captions burned into the picture cannot be turned off, searched or read in braille. The correct form is a separate .vtt or .srt file attached with <track kind="captions">.
Transcript
The full text of the recording as a page or a file beside the player. The only form accessible to a deafblind user — and, incidentally, the only one a search engine can index.
Audio description
An extra narration track describing what cannot be heard: who entered, what the chart showed, what appeared on screen. The cheaper form, allowed at level A: a full text alternative describing the sequence.
Automatic YouTube captions without correction do not satisfy 1.2.2 — the standard speaks of correct captions, not of an attempt at them. A sane process: the machine writes the first pass, a human fixes names, numbers, punctuation and line breaks. That is usually a quarter of an hour per ten minutes of footage.
The player is an interface too
- Every control has a name: “Play”, “Mute”, “Captions” — not just a glyph.
- Full keyboard operation, including the progress slider and the volume control (2.1.1).
- Focus does not get trapped inside the player’s
<iframe>(2.1.2). - No autoplay with sound (1.4.2) and nothing flashing more than three times a second (2.3.1).
- The browser’s native
<video>gives you all of this for free; a custom player has to be tested with a screen reader.
PDF: clause 10, the one everyone forgets
EN 301 549 does not stop at the web page. Clause 10 applies the same criteria to documents that are not pages: PDF, DOCX, spreadsheets, slide decks, e‑books. Terms, a price list, a complaint form and a manual are part of the service — if they are inaccessible, the service is inaccessible, however impeccable the site around them may be.
A scan is not a document
A PDF produced by a scanner is a picture. A screen reader will not read a single word of it. The fix: OCR with correction, or better, regenerating the file from its text source.
Tags and reading order
An accessible PDF is “tagged”: it has a structure of headings, lists, tables and paragraphs, and the reading order matches the visual one. In a PDF editor you check this in the tag panel and the reading-order view; in Word, by using heading styles instead of bold text.
Image alternatives and tables
Every graphic that carries meaning has a description; a chart without one is a blank space. Tables have header rows marked; a table used for layout is a mistake here exactly as it is in HTML.
Language, title, bookmarks
Document properties: the language (so a synthesiser does not read Polish as English) and the title (so the reader does not announce invoice_final_v3.pdf). Anything longer than a dozen pages needs navigation bookmarks.
Forms in PDF
Fields need labels and a sensible tab order, and the document must not be secured in a way that blocks assistive technology. A complaint form to be printed and posted does not replace the channel required by art. 37.
PDF/UA
ISO 14289 sets out the technical requirements for an accessible PDF. Checkers can validate a file automatically — with the same caveat as for web pages: no errors in the validator is not the same as a document somebody can actually read.
The cheapest PDF fix is usually not having one: terms, a price list and a manual published as ordinary web pages are accessible by definition, work on a phone and get indexed. Keep PDF where a signature or printing is genuinely needed.
E‑books and media services
The act names e‑books and media services among the covered services (art. 3(2)) and adds requirements for them (art. 13–17): passing on accessibility metadata, no protection measures that block assistive technology, an accessible player, and a guide to the offering. The practical minimum for a publisher is EPUB 3 with correct structure and an accessibility declaration in the metadata — a file a reader can read and that says what it can do. Sector detail lives on the Other services page.
How to check your own
Inventory the media and the files
Every video, podcast and downloadable file on the site. There are usually three times more than marketing remembers.
Mark what is missing on each
Captions? Transcript? Audio description? Tags in the PDF? That is your list of non-conformities for the art. 32(1) documentation.
Fix in order of traffic
What most people watch and download first, plus anything that is part of buying or complaining. A decade-old archive comes last and may legitimately be recorded as an exception with a date.
Our analyzer detects video with no caption track and sound that starts on its own — the rest of this list is editorial work no tool will do for the editors.