@Jakob Rosin Come to think of it, ideally, I'd provide individual image descriptions with text transcripts in all languages that appear in the image myself.
This way, if I have a sign in French, English and German, I can add a verbatim transcript of the French text to the French description, a verb...