Generic5/3-Regular read out loud backslash

[MedBooster]

IDK what it is, but "backslash" is read out loud randomly for some letters, for Y among others it seems. For the second not the first line in the PDF, I guess it depends on what "font" it is, maybe this is a bug that needs to be fixed? Or maybe this is an issue with how files get shared from iOS to windows...
iOS Version 14.6 Quartz PDFContext
Check out the font type in this PDF
I'm not able to search through this kind of document either....
any idea on how to convert this nonsense font to something "searchable", as well as "readable" for the read-out-loud-tool?

image.png

backslash glitch read out loud 1.09.2024.pdf

edit: I guess for searching you do OCR with "searchable image" ticked, but in that case the read-out-loud-tool will still read it incorrectly I assume.

[MedBooster]

As well with the Generic3-Regular font, when I select text and use the bookmark shortcut to create a bookmark containing the selected text, all spaces are removed, perhaps that is something you could fix as well.

[Daniel - PDF-XChange]

Hello,

From what I can see here, this is a bug with the embedded font, which has incorrect character glyph mappings to the actual in use characters. You can see this in the content pane:

image.png

Or when changing the font in the document to a locally installed and properly formed font, like the locally installed Arial:

PDFXEdit_1Lr7WFJWb8.gif

I can pass this along to our Dev team, but I expect there is little to offer from our end. At this time, the best way to quickly fix this, will be to perform the Enhanced OCR (requires a Plus license) on the document, intentionally reprocessing the existing text on the page, to generate a new text layer with proper fonts:

image(1).png

Kind regards,

[MedBooster]

Will then the read-out-loud-tool read the text produced by OCR instead as well?

[Daniel - PDF-XChange]

Hello, MedBooster

Of course, Read out loud reads the actual text data, not the glyphs on the screen (otherwise the read out loud tool would need to leverage OCR in the first place (which would likely prevent this issue, at the cost of making read out loud, extremely slow). The Enhanced OCR engine, when set with the options I highlighted as they are in my screenshot, will replace the old damaged text with new properly formed text data.

Kind regards,