Yesternight I went looking for Optical Character Recognition (OCR) apps.
For the jargon wary and the jargon weary, OCR is the process by which a computer converts a scanned image of some text to actual editable, malleable computer text. It’s a useful tool to have on hand, especially if your work (or play) deals with anything literary or historical. Like speech recognition software, its also a bit magical.
I wanted to transcribe a few noteworthy pages of a novel, to paste into my Commonplace Book. Rather that wait a few hours until I could use my office facilities to scan and convert the text, I sought recommendations online and did a search of Apple’s App Store.
Re: previous tweet, yes I know OCR *software* exists but I want an app for my phone that does it now now now
— robertsharp59 (@robertsharp59) October 7, 2013
I found a few free options in the App Store. To test the software, I used a piece of text from Neal Stephenson’s Snowcrash. I am well aware that the image is not optimised for OCR – the text is slightly skewed and the contrast could be brighter. But by necessity, a mobile OCR app will be used ‘in the field’ where more careful alignment and lighting is not available. So I think it is an appropriate example to road test the various options. I used my iPhone 4S running iOS 6.1.
IRISCloud is a free app, but to use it you need to sign-up for an account and buy ‘freemium’ packages. I didn’t.
Skootch OCR does not appear to offer any features or control. You simply select the image of the text you need scanning, and the app does the rest. Rather than show you the text immediately it offers the option to e-mail the text to someone (or yourself).
The output was garbage.
.= ] 'QW-Â»~ lQ`Â§","* I .,,Â¢Â»,..Â».v,Â»~Wmn.W,~,,,,M~ W
Image to Text does two weird things. First, it displays any image you choose in the portrait orientation, even if the original is landscape. That is very confusing. Then, it insists on e-mailing the results rather than simply displaying them on the screen for the user to copy-and-paste. The output of the section I requested was near perfect. However, because there is no feature to select precisely which part of the image I wanted to transcribe, I also received more text than I needed. At the top of the page, where the page bowed and the image was distorted, I received increasingly illegible text.
The LEADTools OCR app does offer the user the opportunity to select which text they wish to scan. However, when the user drags a selection box, the OCR scan begins immediately, leaving no room for error. Only the most dexterous draggers-and-droppers would be able to take advantage of that functionality. However, once the text has been selected, the results were perfect. The app offers the user a choice of whether to copy the text to a clipboard, or to e-mail it somewhere.
OCR Scanner offers 5 free daily scans. You have to pay if you want more. It offers no chance to select which area of an image to scan. Its an academic point to note that omission, however – after showing a rotating progress icon for a bit, the app went back to its original state. The 5 credit counter had not depleted, and no recognised text was presented!
OCR Scanning has an extremely confusing text selection tool. The box has a hard green line, and a thicker, slightly transparent region. My first attempt at scanning missed much of the text, because I had allowed some of that transparent green zone to intrude over the letters. On my second attempt, the results were very disappointing. Just a few sentences were scanned and there were many errors.
Y.T.’s rnorn works in Fedland. She has p8IK€(1 ner urue can ll! nu; own little numbered slot, for which the Feds require her to pay about ten percent of her salary (if she doesn't like it she can take a taxi or walk) and walked up several levels of a blindingly lit reinforced-concrete helix in which most of the spaces-—the good spaces closer to the surface--are reserved for people other than her, but empty. She always walks up the center of the ramp, between the rows of parked cars, so that the EBGOC boys won't think she's lurking, loitering, skulking, malingering, or smoking. ‘ * - -- -r L..- i...;A:m. 4... 5“
In my opinion, English OCR has the simplest and best interface. It does offer the option to select an area of the image to scan, and allows you to rotate the image too. The app presents a simple progress message, counting the seconds, so you know the thing is working. The results of the scan were completely accurate, and the text was presented in a simple format for me to copy-and-paste. This is the app I will use in future.