Can you use Windows' built-in OCR? Yes, in Read
Windows has an OCR engine built in, and Read by Zaweba uses it as one of its two engines, beside Tesseract, choosing between them for each job.
Windows does have a text recognition engine built in, but there's no obvious place in Windows to point it at a picture. Read by Zaweba gives you that place. It uses Windows OCR as one of its two engines, alongside the bundled Tesseract 5 engine, and its Auto setting picks the better one for each job.
Windows OCR is part of Windows itself, the same engine various Microsoft features use behind the scenes. It's quick and it's already on your PC. What it lacks on its own is an app for ordinary people: somewhere to drop a picture, check the result and save it.
What Windows OCR is good at
It's fast. For a single picture, a photo, a screenshot, something pasted from the clipboard or a screen grab, it gets the words back quickly. It reads the languages you have installed in Windows, one language at a time.
There are things it doesn't do. It gives no confidence for each word, so Read can't shade the uncertain ones amber when Windows OCR did the reading. It doesn't do the kind of layout analysis that columns and tables need, and it can only read languages Windows has.
Why Read pairs it with Tesseract
Tesseract 5, which comes inside Read, fills those gaps. It analyses layout, gives a confidence for every word, supports more than 40 languages and can read several at once, such as English and Welsh together. It's the engine Read uses for columns, searchable PDFs, tables, multi-page documents, and any language Windows lacks.
Auto does the choosing. Quick single pictures go to Windows OCR when Windows has the language; the bigger, more structured jobs go to Tesseract. Above the text, Read shows which engine it used and why, so the choice is never hidden. If you want a second opinion, Read again with Tesseract is offered in the Read menu.
How the two compare
On a clean A4 letter drawn at 300 DPI, Read's tests measured 99.53 percent word accuracy for Tesseract and 98.60 percent for Windows OCR. Both are good on clean print. The practical differences are the ones above: Tesseract's word confidences make checking easier, and its layout analysis handles columns and tables.
Clean-up matters more than the engine for phone photos. A photo of the same letter taken at an angle with shadows scored the same as the scan after Read's clean-up, and Tesseract on the raw photo with nothing done to it read none of it at all.
Using it day to day
The place you'll meet Windows OCR most in Read is the screen text grab. Press the hotkey, Win+Shift+T unless another program already uses it, drag round some text on screen and it's copied a moment later. Quick, single, small: just what Windows OCR is quick at. Read enlarges the grabbed area to twice its size first, because screen text is small.
To read a language with Windows OCR, Windows needs that language installed. If it doesn't have it, Read simply uses Tesseract, and you can download Tesseract language packs from the Languages tab when you need them.
Private either way
Both engines run on your PC. Neither sends your pictures anywhere, and Read by Zaweba only goes online to fetch a language pack you've clicked to download.
