So AI-based, offline image descriptions are coming to NVDA. Apparently no follow-up questions or longer descriptions for now. The descriptions or captions are also initially in English.
github.com/nvaccess/nvda/pull/…
@NVAccess
#Accessibility #NVDA
in reply to Sean Randall

@cachondo `ui.browseableMessage` is extremely basic, and just lets you choose which HTML to display in the browse mode document window. It doesn't e.g. include a JavaScript bridge so that code inside the document can communicate back with the caller.

I suppose you could run an HTTP server inside the calling code, and include a form in the HTML document that POSTed to it. I'm not sure if that would be disallowed by some sort of security policy, or whether the HTML renderer being used is very modern. @amir @tspivey