I wonder if there’s a way to reverse engineer the words based on the size of these blackouts. If they didn’t add a random variation in size to the black box, then it probably just lines up with the selectable text component iself. Meaning this size could be used to narrow down within the width of the text within the font and how it’s rendered.
Edit: it seems the height for “don’t” (doesn’t have a decender) is the same the height as one with a decender. So provably only the width. The size of a space character is the same throughout a line, so after trying to imagine the “don’t” in that sentence, it seems it’d be right up against the left and right side of that box.
Perhaps a text transformer could be introduced to narrow it down further, or at least present things on statical likelihood
I’ve seen some people doing exactly that. Most of these are in 12pt Times New Roman, so you can literally just overlay a text box lined up with the surrounding text and try known names until everything matches.
On top of that, some of the redactions are done with software that performs the font smoothing / anti-aliasing after drawing the black boxes, which can cause some letters to leak out a little, which can also be matched to known names
I wonder if there’s a way to reverse engineer the words based on the size of these blackouts. If they didn’t add a random variation in size to the black box, then it probably just lines up with the selectable text component iself. Meaning this size could be used to narrow down within the width of the text within the font and how it’s rendered.
Edit: it seems the height for “don’t” (doesn’t have a decender) is the same the height as one with a decender. So provably only the width. The size of a space character is the same throughout a line, so after trying to imagine the “don’t” in that sentence, it seems it’d be right up against the left and right side of that box.
Perhaps a text transformer could be introduced to narrow it down further, or at least present things on statical likelihood
I’ve seen some people doing exactly that. Most of these are in 12pt Times New Roman, so you can literally just overlay a text box lined up with the surrounding text and try known names until everything matches.
On top of that, some of the redactions are done with software that performs the font smoothing / anti-aliasing after drawing the black boxes, which can cause some letters to leak out a little, which can also be matched to known names