The add-on provides a description of photos, for example, a 23-year-old girl with blonde hair smiling, text recognition, text translation into another language and reading QR codes.
Open NVDA menu, Preferences, CloudVision Settings.
For the navigator, just set it to the desired element, press the combination and the add-on will recognize it. For a file, select it in Windows Explorer without opening it, immediately, without opening the file, press the combination. Only JPG, PNG, GIF are supported. I did not add a PDF, since recognition can last 40 minutes or longer.
The idea and code for the selected files is taken from the "Nao (NVDA Advanced OCR)" add-on.