I can definitely say that OCR engines in general have improved more
substantially than I can even express over the last several decades.  You
can have some very, very crappy originals and still get remarkably accurate
results.  For those of us of a certain age where mimeographed copies were
later photocopied you get these things that have little dots scattered all
over the place and blotches here and there.  When I've OCRed these (or image
PDF versions of these) in recent years the results have been nearly perfect,
which is saying something.

I also just tested the Canon IJ Scan Utility with NVDA and all of its
buttons are accessible and, given how little its changed over the years, I'd
expect that the underlying functions for each, when they require user
interaction, will be so as well.  For single page scanning to PDF all you do
is hit one button and you're done.

--

Brian *-* Windows 10 Pro, 64-Bit, Version 1909, Build 18363

*Science has become just another voice in the room; it has lost its
platform.  Now, you simply declare your own truth.*

~ Dr. Paul A. Offit, in New York Times article, How Anti-Vaccine Sentiment
Took Hold in the United States (
https://www.nytimes.com/2019/09/23/health/anti-vaccination-movement-us.html
) , September 23, 2019