IDEA: OCR & paste

[alien]

I think of a tool to select and copy/paste text from image files.

[iorange28880]

That's a neat idea, to have OCR built into CH+S.

[mouser]

there is a fantastic shareware to do this kind of stuff - it's called Kleptomania:
http://www.structurise.com/kleptomania/

[kfitting]

Problem is Kleptomania (and every other software I saw to do this type of thing) cannot OCR a picture displayed on screen.  This would be an awesome feature... though no doubt harder.

Kevin

[rjbull]

Problem is Kleptomania (and every other software I saw to do this type of thing) cannot OCR a picture displayed on screen.

kfitting,

I just googled Screen Desktop OCR.  The features list includes the following - note particularly the section in parentheses in the first claim, BMP and JPG images:

Recognizes text in any type of document (HTML, Word, Notepad, PDF, BMP or JPG image) or in any part of the screen and saves it to clipboard.

Grabs texts that cannot be saved or printed

Recognizes texts in any language that cannot be saved or printed

Works with all American and European character sets, including the Central-European and Cyrillic

Captures "hidden" text ignoring blocking scripts on Internet pages

Sanity check: I have not tried it myself.  $29.95 shareware.

[kimmchii]

ABBYY screenshot reader can capture everything.

[rjbull]

ABBYY screenshot reader can capture everything.

Does "scrape text" from othewise difficult things, like OCRing text in images?

Do you have a URL?

[kimmchii]

unfortunately, it comes with ABBYY FineReader 8.0, not a separate program.

yes it can easily scrap text from images as long as it's legible.

you can send me the image and i will try to OCR it for you, see how well it can read.

[kfitting]

I've tried Screen Desktop OCR... it seems like a Kleptomania clone to me.  There are about 3-4 programs that claim to do the same thing and do it almost exactly the same way!  (Even down to the program icon... Top OCR is another name that sounds familiar, but I could be wrong)  The problem is, they dont seem very reliable.  They do great with text in a window, but pictures must need to be perfect cause I've never gotten a picture to work.  I tried them on AutoCAD and got a bunch of gibberish.  I don't know, they just need more refinement or something. 

I like ABBYY's claim to be able to use a digital camera as a scanner, that's awesome... but the price tag is, needless to say, pricey! 

Kevin

[rjbull]

I've tried Screen Desktop OCR... it seems like a Kleptomania clone to me.  There are about 3-4 programs that claim to do the same thing and do it almost exactly the same way!  (Even down to the program icon...

I was astonished to see the same icon.  Dog should not eat dog.

I like ABBYY's claim to be able to use a digital camera as a scanner, that's awesome... but the price tag is, needless to say, pricey! 

Eye-watering 

Going to take up kimmchi's offer to see if ABBYY will read a typical image?

<later edit>        TopOCR: yes, you're right, with "camera as scanner" apparently the main purpose of the program.

Digital Camera OCR

Can your camera or smartphone replace a scanner for document capture with OCR? The answer is a resounding YES! TopSoft, Ltd, a leading developer of OCR and image processing software has created an application specifically for OCR data capture with digital cameras and smartphones called TopOCR.

I notice both Kleptomania and ScreenOCR are about 400Kb to download.  TopOCR is ten times that, about 5.4Mb.  "Shareware" in that it only runs ten times rather than {x number of days} before needing registration.

However, I have less personal need of image-reading than of things like extracing bits of Outlook message headings lists, and I imagine Kleptomania and ScreenOCR would do that adequately.

[mouser]

If someone is willing to write up a little summary of the pros and cons of these different tools, we'll see if we can't contact them and get a discount for members.

[app103]

Maybe some day if mouser is feeling adventurous, he might add something like this to Screenshot Captor?

Take a screenshot and then use OCR to scrape text from that screenshot, and save the text, rather than the image.

[mouser]

i actually looked into this using an open source ocr tool, but it did such a bad job of ocr that i ditched the idea.
do remember that screenshot captor has lots of functions for adding 3rd party commandline tools as menu items so if someone has an ocr tool that works via commandline, you could interface it.

[murple]

I think this is a fantastic idea. I've been wanting to make a small AutoHotkey script for some time which is a simplified version of this. It would just pop-up a transparent, non-intrusive inputbox, allow you to type into it and when you hit enter it disappears and copies the text to the clipboard. I do this sort of thing many times everyday, but since I haven't made the program I usually end up hitting Win+R to type into the Start -> Run box, then do Shift+Home and Ctrl+C, Esc. Anyhow, I just wanted to express my enthusiasm for this project. Too bad the opensource OCR package Mouser found wasn't up to the task.

[rjbull]

would just pop-up a transparent, non-intrusive inputbox, allow you to type into it and when you hit enter it disappears and copies the text to the clipboard.

Murple,

Unless I'm missing something, this doesn't look like OCR... if you just want to put text into the clipboard, there's lots of ways of doing it.  E.g. using Mouser's own Clipboard Help + Spell, you call up the main window, Edit, Insert New Clip.  ClipCache has similar.  Or, from a command line, try Horst Schaeffer's ClipText.

[rjbull]

Problem is Kleptomania (and every other software I saw to do this type of thing) cannot OCR a picture displayed on screen.  This would be an awesome feature... though no doubt harder.

kfitting,

I just noticed that the beta of the full (payware) version of EverNote 2.0 claims to be able to do this:

AIR-search in images: Search photos and graphics stored in your notes to find embedded typed or handwritten text. Use it for snapshots of whiteboards, product labels, in-store packaging and pricing, business cards, trade show badges, and more.

[kfitting]

Looks pretty cool... I played with it a little and that is a nifty feature to have.  Unfortunately, Evernote is a huge paradigm shift for me and I don't see myself using it (though I can see where it has a lot of power!). 

As far as OCR stuff is concerned, I wish you could export the text to an RTF file.  That's what I'd really like... to be able to take a picture (or a screenshot) and turn it into editable text.  But, thanks for the heads up on that.  At least someone is developing along these lines!

Kevin

[murple]

rjbull,

Unless I'm missing something, this doesn't look like OCR...

You're entirely right, that is not OCR (although one could -stupidly- argue that it's OCR performed by the user as opposed to the computer :). I would love to have a program which would allow me to select part of my screen, perform OCR on it and copy the recognized text to the clipboard. Unfortunately, coding anything involving OCR is lightyears out of my league.

Instead then, I'd like to conjure up a tiny application which allows you to "write directly to the clipboard" in the easiest, fastest most intuitive and least intrusive way. I'd call it ClipWrite.

Thank you for the program suggestions but they weren't quite what I had in mind.

[murple]

ClipWrite isn't very good yet, mostly because it's slow because the code sucks and it's still buggy. But in functionality, this is what I had in mind:

#singleinstance force
;Run the script, then enter Win+C and type your text. Hitting the enter key, will hide the window and copy what you wrote to the clipboard. Ready to be pasted elsewhere.
coordmode, mouse, screen
;The codeline Enter:: means that hitting enter will close the program and copy the text. This makes it very quick to close, but it disables your enter key when elsewhere when the program is running. What is the workaround for this? I want the code in the Enter:: section to be executed only if the gui


#z::
gosub, BuildGui
return



Enter::
GuiControlGet, MyEdit
if ErrorLevel = 0
  {
  clipboard = %ControlText%
  Gui, destroy
  SetTimer, UpdateGUI, Off
  }
return

#q::ExitApp


BuildGui:
CustomColor = EEAA99  ; Can be any RGB color (it will be made transparent below).
Gui, +AlwaysOnTop +LastFound +Owner  ; +Owner prevents a taskbar button from appearing.
Gui, Color, %CustomColor%
Gui, Font, s20
Gui, Add, Text, w500 vMyText c800080, XXXXX YYYYY  ; XX & YY serve to auto-size the window.
Gui, Add, Edit, vMyEdit
GuiControl, Move, MyEdit, x10000 y10000 
;GuiControl, hide, MyEdit


; Make all pixels of this color transparent and make the text itself translucent (150):
WinSet, TransColor, %CustomColor% 200
Gui, -Caption  ; Remove the title bar and window borders.


SetTimer, UpdateGUI, 50
  Gosub, UpdateGUI  ; Make the first update immediate rather than waiting for the timer.
return

return


UpdateGUI:
MouseGetPos, MouseX, MouseY
msgWidth = 570
msgHeight = 120
SysGet, MonitorWorkArea, MonitorWorkArea, %A_Index%
maxRight := (MonitorWorkAreaRight - msgWidth)
maxBottom := (MonitorWorkAreaBottom - msgHeight)


;msgbox, msgWidth:%msgWidth%`nmsgHeight: %msgHeight%`n`nMouseX: %MouseX%`nMouseY: %MouseY%`n`nMonRight: %MonitorWorkAreaRight%`nMonBot: %MonitorWorkAreaBottom%`n`nMaxRight: %maxRight%`nMaxBot: %maxBottom%

if(MouseX > maxRight)
{
MouseX = %maxRight%
}

if(MouseY > maxBottom)
{
MouseY = %maxBottom%
}

Gui, Show, x%MouseX% y%MouseY%
;GuiControl,, MyText, x%MouseX% y%MouseY%
GuiControlGet, ControlText, , MyEdit, Text
GuiControl,, MyText, %ControlText%
return

Could be used together with Jim Biancolo's PlainPaste as a sort of clipboard extension.

[rjbull]

one could -stupidly- argue that it's OCR performed by the user as opposed to the computer

I thought that was traditionally known as "reading" 

I would love to have a program which would allow me to select part of my screen, perform OCR on it and copy the recognized text to the clipboard.

That's exactly what Kleptomania does.

I'd like to conjure up a tiny application which allows you to "write directly to the clipboard" in the easiest, fastest most intuitive and least intrusive way. I'd call it ClipWrite.

Good luck with that, but, you could always keep a small fast editor like TED Notepad running in the background?

[kartal]

kleptomania`s success rate is not very high.I have tried it in xyplorer and gave me gibberish for example.

one could -stupidly- argue that it's OCR performed by the user as opposed to the computer

I thought that was traditionally known as "reading" 

I would love to have a program which would allow me to select part of my screen, perform OCR on it and copy the recognized text to the clipboard.

That's exactly what Kleptomania does.

I'd like to conjure up a tiny application which allows you to "write directly to the clipboard" in the easiest, fastest most intuitive and least intrusive way. I'd call it ClipWrite.

Good luck with that, but, you could always keep a small fast editor like TED Notepad running in the background?

[yksyks]

I would love to have a program which would allow me to select part of my screen, perform OCR on it and copy the recognized text to the clipboard.

I'm using for this purpose JOCR. Not very new, not always 100 % reliable, but free, simple and does what it's supposed to.

[kartal]

It looks like JOCR requires Msoffice to be installed? I do nto have it and it gives me error.

I would love to have a program which would allow me to select part of my screen, perform OCR on it and copy the recognized text to the clipboard.

I'm using for this purpose JOCR. Not very new, not always 100 % reliable, but free, simple and does what it's supposed to.

[yksyks]

It looks like JOCR requires Msoffice to be installed? I do nto have it and it gives me error.

You're right, I completely forgot it, sorry. From the website:

JOCR requires Microsoft Office 2003 or higher version.

Anyway, for MS Office users it's quite useful tool.

[rjbull]

kleptomania`s success rate is not very high.I have tried it in xyplorer and gave me gibberish for example.

It doesn't work on very small fonts.  Could that be the problem?