Am Tue, 29 Apr 2025 13:13:39 -0400 schrieb Matthias Weber:
Dear All,
I am starting a new thread as requested.
I can confirm that
the following should work with current latest (from yesterday, LuaMetaTeX 2.11.07 20250427 + ConTeXt LMTX 2025.04.28 14:29):
\enabledirectives [backend.usetags=testing] \setuptagging[state=start] \setupstructure[state=start] \setupbackend [format=PDF/A-3a] \setupbackend[format=pdf/ua-1] \setupexternalfigures[location=default] \starttext \externalfigure[hacker][label={this is an image}]
\startparagraph a \stopparagraph \stoptext
produces a PDF that passes the accessibility check in the current Adobe Acrobat. In particular, it passes the alternate text checks. Hovering over the image shows the alternate text in Acrobat. However, VoiceOver (on the Mac) does not read the alternate text;
The pdf produced by context maps every structure to NonStruct, that means there is no tagging structure, it only fools the accessibility check. You can and should not claim that this an ua-1 file. You can check the structure by uploading your pdf to https://texlive.net/showtags, that will produce an xml from the structure. I don't know if the missing structure is the reason that VoiceOver fails, as I'm on windows, but you could try with a LaTeX document. Go to https://texlive.net/ngpdf and compile there this example: \DocumentMetadata{tagging=on} \documentclass{article} \usepackage{graphicx} \begin{document} some text \includegraphics[alt=an image with a duck]{example-image-duck} \end{document} After the compilation you can download the pdf and compare the reading. I made a fast check with Adobe + NVDA: it reads the actual text of your image, but it doesn't announce that there is graphic. So for the context file you get simply this is an image while the LaTeX file is read as graphic an image with a duck. -- Ulrike Fischer http://www.troubleshooting-tex.de/