New Free OCR Tool

Daniel Gaskell has recently released ScreenOCR, a front-end for the Tesseract engine that lets you instantly OCR anything visible on your screen with a single click. After playing around with it, ScreenOCR gets my official stamp of approval: it’s the easiest way yet to OCR text so that it can be copied into your debate template. The download is available online for free along with some suggestions for using the software. One hint that I’ll echo: make sure you zoom in so that the text is very large before using ScreenOCR. Once I figured this out, OCRing text from Google Books became a snap. Thanks to Daniel for making this software available to the debate community.

11 thoughts on “New Free OCR Tool

  1. danielgaskell

    Also, if you're getting an error while installing to Program Files, try selecting a different folder or running the installer as an administrator. (Windows doesn't always ask for elevated privileges for some reason – I should probably figure out why…)

  2. Antonucci23

    Bill (or anyone else) –

    How do you find this compares to ABBYY Finereader specifically? I'll try both, but I'm curious as to your take on a specific product comparison.

    1. Bill Batterman Post author

      I don't have much experience with Finereader — I have been using Omnipage Pro on the Mac for a long time and it has always served me well, but this ScreenOCR program makes it so much easier to just grab a couple paragraphs or pages. My old process involved downloading images, creating a PDF, and running that through Omnipage… lots of work for just a couple cards.

      I'm also interested in any side-by-side comparisons. Anyone used both? Or other programs, for that matter?

      1. danielgaskell

        Okay, I just did a few tests with the trial version of ABBYY Screenshot Reader (which, by the way, you can actually buy separately from FineReader for only $10 here: http://www.abbyyusa.com/shop/screenshotreader/ – I didn't know this before.)

        The first difference I noticed is that ABBYY doesn't have a global hotkey – you have to open the window from the system tray every time you want to convert something. Inconvenient, but minor.

        **** EDIT NOTE: The tests below are for an earlier, less accurate version; see the end of the comment for updated results.

        At higher zoom levels, the difference is slight. The Tesseract engine misread a period, which ABBYY doesn't, but gets everything else right:

        * ABBYY, zoomed in:
        Presidential framing sets the foreign policy tone for the American public. One
        would be hard pressed to argue that the president's framing of foreign policy
        events has no effect on the American people. The main question is what type of
        framing and rhetoric are effective in persuading the American public to support
        risky foreign policy?

        * ScreenOCR, zoomed in:
        Presidential framing sets the foreign policy tone for the American public. One
        would be hard pressed to argue that the president's framing of foreign policy
        events has no effect on the American people, The main question is what type of
        framing and rhetoric are effective in persuading the American public to support
        risky foreign policy?

        At small size, however, the difference is much more noticeable. ABBYY makes one error; Tesseract produces almost complete gibberish:

        * ABBYY, no zoom:
        Presidential Training sets the foreign policy tone for the American public. One
        would be hard pressed to argue that the president's framing of foreign policy
        events has no effect on the American people. The main question is what type of
        framing and rhetoric are effective in persuading the American public to support
        risky foreign policy?

        * ScreenOCR, no zoom:
        P.¢s|a¢m|.~| ammmg ms um mug.. Pauly mm form Amfnm puhnc om
        would b¢ mm pmsm m argue nm mf pf¢s|a¢..¢'s fmming of fnmgn pulmy
        ¢f¢m< rm ».» mm .m nw Amcnm ,1¢<»p|¢_ 'rm mm qu¢=n.m N www we of
        fmmmg and flmonf me =¢mw¢ m p=mmamg me Ammm pubnn 1° suppon
        mky rumgn polmyv

        I tried a few other pages, with similar results. ABBYY makes enough errors at normal size that it's worth zooming in anyway, though.

        My takeaway? If you've got ten bucks lying around and don't mind the lack of a hotkey, ABBYY is better. If you're a cheapskate, just download ScreenOCR and remember to zoom in.

        **** UPDATE: Version 1.1 improves conversion significantly. Google Books with no zoom now converts about as well as ABBYY:

        Presidential framing sets the foreign policy tone for the American public. One
        would be hard pressed to argue that the president's framing of foreign policy
        events has no effect on lhc American people. The main question is what type of
        framing and rhetoric are effective in persuading the American public to support
        risky foreign policy?

    2. danielgaskell

      Finereader's screen-capture tool is a lot more sophisticated – it will do advanced page layout analysis, etc. Tesseract (what this uses) pretty much just chucks out a block of text. Finereader also has a slightly lower error rate, especially on smaller text (the difference isn't much if you zoom in.)

      The main advantages of ScreenOCR are a) it's fast and lightweight (instant conversion), and b) it's free.

  3. timalderete

    First – Most Excellent tool. Second – a Question. When fully zoomed, is there a way to capture more than what is on the screen? Is there a way to have the screen scroll down if I hold down the Mouse Left and drag it below the screen? I've tried a couple of times, and it doesn't seem to scroll. Not a big deal if I am just copying a paragraph or card, but is it possible? Thanks.

  4. danielgaskell

    In a sudden burst of fridge logic, I realized that the conversion of low-resolution text could be significantly improved by… just increasing the resolution of the image file. Shock!

    The difference is amazing. For Google Books without any zoom, it goes from this:

    P.¢s|a¢m|.~| ammmg ms um mug.. Pauly mm form Amfnm puhnc om
    would b¢ mm pmsm m argue nm mf pf¢s|a¢..¢'s fmming of fnmgn pulmy
    ¢f¢m< rm ».» mm .m nw Amcnm ,1¢<»p|¢_ 'rm mm qu¢=n.m N www we of
    fmmmg and flmonf me =¢mw¢ m p=mmamg me Ammm pubnn 1° suppon
    mky rumgn polmyv

    To this:

    Presidential framing sets the foreign policy tone for the American public. One
    would be hard pressed to argue that the president's framing of foreign policy
    events has no effect on lhc American people. The main question is what type of
    framing and rhetoric are effective in persuading the American public to support
    risky foreign policy?

    Anyway, the long and short of it is, there's a new version out. While I was at it, I also added an option to hide the About window on startup, and fixed the encoding issue with curly quotes. Enjoy!

  5. RyanCMarcus

    Google Docs does this for free as well. Just upload an image or PDF file. The algorithm is pretty damn good — Google was behind one of the original open source OCR projects, and they've improved their version drastically. Not sure how it stacks up, but worth taking a look at.

    Source: http://docs.google.com/support/bin/answer.py?answ

    Just to clarify my own comment — yes, you can't use this to instantly OCR things on your screen. But if you screen capture and then upload, this works too. Just an extra step, I suppose.

  6. Scottyp4313nr Post author

    Gaskell,

    I tried installing this on a new computer today and couldn't get it to work. I first changed the folder as you suggested and it would install, but now it opens up and gives me an error message anytime I try and do anything even hit hide. The error is "cannot find path blah blah blah". This is on a 64 bit windows 7 computer

Comments are closed.