PDF does also support alt-texts (and has some other accessibility support). See 14.9.3 Alternate Descriptions in https://opensource.adobe.com/dc-acrobat-sdk-docs/pdfstandards/PDF32000_2008.pdf

PDF does also support alt-texts (and has some other accessibility support).

I think a parallel issue to this one is track down which file types/backends support alt-text and update accordingly. Tom put in support for svg in #21328

I just sent an email to a Swedish researcher asking for some pointers. Let us see if I get some answer. Also, found this paper which seems quite relevant: https://content.iospress.com/articles/technology-and-disability/tad170181 (I have not read it yet, but I guess the references and citing papers can be worthwhile, even if we may not be interested in automation in the first place).

Edit: this may be worthwhile checking as well https://viewplus.com/product/audio-graphing-calculator/

https://dev.exiv2.org/projects/exiv2/wiki/The_Metadata_in_PNG_files

https://iptc.org/std/photometadata/specification/IPTC-PhotoMetadata#alt-text-accessibility

https://exiv2.org/tags.html / https://www.cipa.jp/std/documents/e/DC-008-2012_E.pdf

There are reasonable places to put "alt text" (although IPTC makes a distinction between "ald text" and "image description").

options:

• minimal: nothing (but docs)! currently fig.savefig('...png', metadata={'alt': 'The alt text'}) works (any key works), also works fig.savefig('/tmp/test.svg', metadata={'Description': 'This is alt text'}) , fig.savefig('/tmp/test.pdf', metadata={'Alt': 'This is alt text'}) also works (but warns). Other raster formats do not work (fails in imsave), but that is a matter of sorting out the right invocation to pillow.
• medium: add a alt_text keyword to savefig that is threaded through each of the backends to translate to the correct name for that file format
• medium++: add a alt_text or description attribute to Figure that is the default value threaded through savefig(..., alt_text)

open questions:

• do any screen readers actually look at these?
• if there are conflicting standards, how to we pick with standard to follow?

Other tasks:

• if we do medium++, teach matplotlib-inline to extract the same for the html
• if we do the medium++, teach sphinx-gallery to extract the same
• if we do medium++ have our plot directive respect this