from nvda user guide:

9. Content Recognition
When authors don't provide sufficient information for a screen reader
user to determine the content of something, various tools can be used
to attempt to recognize the content from an image. NVDA supports the
optical character recognition (OCR) functionality built into Windows
10 to recognize text from images. Additional content recognizers can
be provided in NVDA add-ons.
When you use a content recognition command, NVDA recognizes content
from the current navigator object. By default, the navigator object
follows the system focus or browse mode cursor, so you can usually
just move the focus or browse mode cursor where desired. For example,
if you move the browse mode cursor to a graphic, recognition will
recognize content from the graphic by default. However, you may wish
to use object navigation directly to, for example, recognize the
content of an entire application window.
Once recognition is complete, the result will be presented in a
document similar to browse mode, allowing you to read the information
with cursor keys, etc. Pressing enter or space will activate (normally
click) the text at the cursor if possible. Pressing escape dismisses
the recognition result.
9.1. Windows 10 OCR
Windows 10 includes OCR for many languages. NVDA can use this to
recognize text from images or inaccessible applications.
You can set the language to use for text recognition in the Windows 10
OCR settings dialog. Additional languages can be installed by opening
the Start menu, choosing Settings, selecting Time & Language -> Region
& Language and then choosing Add a language.
To recognize the text in the current navigator object using Windows 10
OCR, press NVDA+r.

--
By God,
were I given all the seven heavens
with all they contain
in order that
I may disobey God
by depriving an ant
from the husk of a grain of barley,
I would not do it.
imam ali