Scanned documents

A scanned document is a picture of a document. You can see the words, but a search finds nothing, you can’t select or copy them, and a screen reader has nothing to read. Nixt PDF can recognise the words and can improve the scan itself. Both run entirely on your computer; nothing is sent anywhere.

Recognising text

Recognition reads the words off each scanned page and writes them onto the page invisibly, exactly where they were found, over the picture. The page looks exactly as it did — every mark of the scan is still what you see — and searching, selecting, copying and screen readers start working.

1. On the Protect tab, click Recognise text… (or choose More › Recognise text…).
2. Read what the Recognise text dialog found, for example:
   • All 6 pages are pictures of pages — nothing in this document can be searched or copied.
   • 2 of 10 pages are pictures of pages. The rest already have text.
   • Every page already has text on it. Nothing to recognise.
3. Under Language, choose the language the document is written in. Whatever this machine is set to is chosen first; the other languages listed are the ones your computer can recognise. A page read in the wrong language comes back as words that nearly exist.
4. Click Recognise. It’s dimmed when there is nothing to recognise.

While it works, the button reads Reading…, a progress bar fills, and the dialog says Reading page 3 of 6…. Each page takes a second or two. The dialog can’t be closed until it finishes.

Only pages that are pictures of pages are read. A page counts as one when it carries almost no text, so a page number a scanner stamped on it doesn’t stop it being recognised. Pages that already have text are left alone. If one page can’t be read, the others still are.

When it finishes, the message says, for example, Read 840 words off 6 pages. The document can be searched and copied now; it still looks exactly as it did. — followed by 12 more were too unclear to write down. when some words were left out.

Each recognised page is its own step for Undo.

Words it was unsure of

A word the recogniser wasn’t sure of is left out rather than guessed at: an invisible wrong word would make a search find a page that doesn’t say what was searched for. But these are often the words people search for — names, reference numbers, foreign words — so Nixt PDF shows them to you.

When recognition finishes with unsure words, the Words it was unsure of dialog opens. Each row shows:

• a tick box;
• the page, such as p3;
• the word as the recogniser read it, in a box you can correct;
• how sure it was, such as 62%.

To add words:

1. Tick each word that is right.
2. Correct any word that is nearly right by typing in its box. Editing a word ticks it for you.
3. Click Add 5 (the number ticked). It reads Add none and is dimmed while nothing is ticked.

The message says Added 5 words to the document. To add none of them, click Leave them out.

Improving a scan

Improve the scan… straightens crooked scanned pages, takes out speckle, and sharpens edges.

1. On the Protect tab, click Improve the scan… (or choose More › Improve the scan…).
2. The Improve the scan dialog says how many pages are scans it can improve, such as 3 pages are scans.
3. Choose what to do. All three are on to start with:

4. Click Improve. It’s dimmed when there are no pages to improve or every option is off.

While it works, the button reads Working… and the dialog says Working on page 2 of 3…. The dialog can’t be closed until it finishes.

When it finishes, the message says, for example, Improved 3 pages, straightening 2 of them. The most crooked was 2.4° out., or Cleaned up 3 pages. Nothing needed straightening.

How straightening decides:

• The angle is measured from the lines of text on the page.
• A page less than a tenth of a degree out is left as it is, because every turn softens a page slightly.
• A page more than fifteen degrees out is treated as a page on its side rather than a crooked one, and isn’t straightened. Use Pages › Turn left or Turn right for that.

Which pages can be improved

Improving a page replaces its picture with a corrected one. That is only safe on a page that is a single scanned picture covering the page, so only those pages are offered. A page with real text on it is left alone: improving it would turn searchable, selectable text into a picture of text. If no page qualifies, the dialog says No page in this document is a scan that Nixt PDF can improve.

What comes out is a new picture rather than the original scan. Undo puts the original back, one page at a time.