Capture2text

If you're having trouble with this application, read on, I found a partial solution.

I tried this program several months ago and found that it was incredibly inaccurate. Wrong characters, tons of extra garbage and numbers in places they shouldn't be. I'm just trying to copy columns of dollar-values from websites so I can more quickly add them up. It is much faster and more accurate to simply copy and paste the lists and then compare the copy for accuracy than it is to hand key each entry manually and then compare.

EasyScreenOCR seemed to be 100% accurate every time and worked perfectly. Well, I couldn't get myself to pay the premium for ESO, so 4 months later here I am trying to do this again and I decided to give C2t another try since others seemingly aren't having the problems I had.

Lo and behold, there's a serious oversight in the default configuration of the program (at least on my system when I downloaded it) that led to the problems I experienced. In the settings window, on the "OCR 1" tab is a "whitelist" box. For whatever reason, the default is to have the whitelist enabled but ONLY have numbers 1234567890 in this box. The problem is that the program will then try to turn absolutely everything into one of those numbers! Dollar signs, letters, background graphics, etc. all get turned into numbers which makes the result useless. If you simply hit the three "add ___" buttons to add actual letters, and then add any symbols that may appear in your list (in my case, decimals, dashes and dollar signs... I also added a space for good measure), then do the same capture, it will likely work much better. Also, simply unchecking the "Enable" button for white lists seems to give a similar, much more accurate result (why even have a white list if it just makes the program misread things?). On the page I was trying to capture I found C2t's accuracy with numbers and decimals to be perfect in about half a dozen captures of 10+ lines each. I still have problems with certain letters though. The letters "il" are captured as either '', H or n . All of them will be the same on each capture, but it changes each time I do a capture. It's almost like there's some kind of issue with the resolution or scaling of the capture which makes thin characters actually capture differently depending on where the capture is started.

Overall though, I am very glad I gave this another try. Hopefully there is a fix for the text accuracy, but for numbers it is currently doing exactly what I need. Whether I saved any time fighting with these programs versus just hand keying everything I needed TODAY, I'm not sure, but in the long run it will certainly save me a lot of time.