Matterhorn 05-002 · PDF/UA-1 UA1:7.1-7

Sound annotations: a speaker icon with nothing to read

A Sound annotation carries information with no transcript.

What a screen reader actually does

A deaf or deafblind reader gets nothing from the annotation.

What a Sound annotation is

A Sound annotation is an icon on the page — a speaker or a microphone — that plays a recording when activated. The recording is inside the file, as a sound stream:

33 0 obj
<< /Type /Annot /Subtype /Sound
   /Rect [ 72 700 96 724 ]
   /Name /Speaker                 % the icon: /Speaker or /Mic
   /Contents (Reviewer's comment)  % its description
   /Sound 34 0 R
>>

34 0 obj
<< /Type /Sound
   /R 8000 /C 1 /B 8 /E /Raw      % sample rate, channels, bits, encoding
   /Length 4000 >>
stream … endstream

Its typical use is a spoken comment pinned to a passage: a reviewer’s note, a pronunciation, a recorded instruction beside a form field. PDF 2.0 deprecated Sound annotations in favour of the general media mechanism, but PDF/UA-1 documents are PDF 1.x, and files carrying them still turn up.

A close relative is the Play a sound action, which Acrobat lists among the actions a link, button or form field can run (Adobe: PDF actions overview). No speaker icon then — the sound plays when something is clicked. The protocol’s condition names the annotation, but a reviewer should treat the action the same way.

What the reader gets

A screen-reader user moving through the page reaches the annotation and hears its description — Reviewer’s comment — and can play it.

A deaf reader sees a speaker icon. Clicking it plays something they cannot hear. A deafblind reader on a braille display gets the description and nothing more. If the recording says “the figure in table 3 is wrong, use 4,210”, neither of them will ever know.

That is the gap this condition names. /Contents is the annotation’s description, required by 28-004 so that it is not a nameless object. 05-002 is about its content: what is said in the recording has to be available in another form.

Where it comes from

• Review workflows with recorded comments, where the PDF that went out for comment is the one that got published.
• Language and teaching material with a pronunciation or listening clip beside the text.
• Forms with spoken instructions added for readers who struggle with text — which helps them and, with no text equivalent, shuts out a different group.
• Old interactive documents built when Sound annotations were the obvious tool.

How to check it free

Find them. They are small and easy to miss on the page. This lists every media annotation — Sound included — with its page and description, using Python and the free pikepdf library:

import sys, pikepdf

MEDIA = {"/Sound", "/Screen", "/Movie", "/RichMedia"}
pdf = pikepdf.open(sys.argv[1])
for n, page in enumerate(pdf.pages, 1):
    for a in page.get("/Annots", []):
        if str(a.get("/Subtype")) in MEDIA:
            print(f"page {n}: {a.Subtype}  Contents={a.get('/Contents')}")

To find Play a sound actions, Adobe documents each object’s actions on the Actions tab of its Properties dialog (Adobe: Add actions to PDF elements).

Listen to each and look for the words. For each recording: does the document say the same thing in text — in /Contents, in the passage beside it, in a transcript elsewhere that the description points to? If not, it fails.

Acrobat’s Accessibility Checker covers annotations under Tagged annotations and Tagged multimedia (Adobe: Create and verify PDF accessibility). Both are about tagging; neither listens to anything.

How to fix it free

For a short recording, put the words in the description. Text in the annotation’s description is available to every reader, and a ten-second comment is a sentence or two. In Acrobat Pro, the description of an annotation’s element can be set in the Tags panel: select the <Annot> element, choose Properties from the Options menu and fill in Alternate Text on the Tag tab. Adobe documents that route for link elements (Adobe: Edit document structure with the Content and Tags panels); the dialog is the same for any tag.

For anything longer, write a transcript and put it in the document as ordinary tagged text — after the passage the recording belongs to, or in an appendix — and let the description say where: Spoken instructions for section 2. Transcript on page 9.

Or replace it. If the recording only repeats text already on the page, it is decoration with a play button; say so in its description. If it is a review comment that should never have been published, remove it.

What Taggart does

Taggart has no check for this condition and cannot listen to a recording; the condition does not appear in the findings or the review queue. It does appear on the list of conditions Taggart does not check, which stands between a clean report and the PDF/UA identifier: the identifier is written only after a person confirms the list.

It does check the annotation itself, and this was tested on a generated file. A Sound annotation outside the structure tree is reported as 28-002; with no /Contents and no alternate text on its element, it is also 28-004. The annotation-structure repair is safe and puts the annotation in an <Annot> element, placed by where it sits on the page. The description is not invented: 28-004 stays open until someone supplies one. In the application, that finding is routed to the review queue with the annotation’s region highlighted, where the decision you make about it is recorded — but the queue’s text box is for figure and formula descriptions, so the annotation’s own description has to be written in Acrobat.

And a described annotation is all Taggart can see. On a second generated file, a Sound annotation correctly tagged with the description Spoken instructions, and no transcript anywhere, passed every check. The conformance statement stopped at machine checks passed, with 05-002 listed as not yet confirmed by a person, and the identifier was not written until that confirmation was given. Whether the words exist is your check to make before you give it.

How Taggart handles it

No software can decide this one, and Taggart has no check for it, so it never appears among the findings or in the review queue. It is on the list of conditions Taggart asks you to confirm by hand before it writes the PDF/UA claim into the file — the claim waits until you have.

Where this sits in the standards

Related conditions

• 05-001 Media annotation present, but audio content not available in another formEmbedded media carries information in audio with no transcript or equivalent text in the document.
• 05-003 JavaScript uses beep but provides no other notificationForm validation signals an error only with app.beep().
• 28-002 An annotation is not a direct child of an Annot structure elementNon-widget, non-link annotations must sit inside an Annot structure element so they have a place in the reading order.
• 28-004 An annotation has neither a Contents entry nor an alternative descriptionEvery non-widget annotation needs an accessible description, either in its own /Contents or in the /Alt of the structure element that encloses it.

Other conditions in checkpoint 05

Checkpoint 05 covers sound. These are its other failure conditions; the ones without a link are in the rule set but do not have a written page yet.

• 05-001 Media annotation present, but audio content not available in another formEmbedded media carries information in audio with no transcript or equivalent text in the document.
• 05-003 JavaScript uses beep but provides no other notificationForm validation signals an error only with app.beep().

This condition needs a person’s judgement, so there was nothing for a tool to check. Written from ISO 14289-1 and the Matterhorn Protocol on 2026-10-02. The Acrobat steps follow Adobe’s documentation and were not walked through by hand. Last updated 2026-10-10.