automatic ocr on elements without description.

[fernando-jose-silva]

Is your feature request related to a problem? Please describe.

Describe the solution you'd like

Voice Over for iPhone automatically detects when a control in focus has no label, and possibly reports xxxx
Where xxx is the text obtained from an OCR.
The suggestion would be to implement a feature similar to NVDA, when the reader detects a control without a label, it tries to run the OCR and inform the user of the text obtained if possible.

Second suggestion:
Voice Over even has a basic feature to detect control shapes, such as right arrows, and informs the user of possibly going back.
If NVDA could recognize commonly used icon shapes and describe them, it would be interesting.
I am not suggesting photo recognition here, because I think we would be talking about artificial intelligence, a subject for another much more complex call.

The most plausible suggestion, if any, would be the first.

Describe alternatives you've considered

Additional context

[Adriani90]

Jaws has this feature too in its 2025 update, works only when using the Jaws cursor or the invisible cursor.
This would be a great improvement in NVDA. I guess it could work at least for navigator / review cursor.
Cc: @jcsteh is this technically possible at all? Voice over does this on unlabelled buttons at least. I didn‘t test with Jaws in detail yet.

[Emil-18]

@Adriani90 I haven't tested this feature in JAWS either, but from the what's new page, it seams like it is just continuous OCR for JAWS, wich NVDA allready has. Voiceover can detect things such as the role and states of objects, e.g that the recognized text is the label of a check box, and that the checkbox is checked.