I have just posted version 1.07 of Capture2Text.

Download Capture2Text v1.07 via MediaFire (source code is included)

(Deposit Files Mirror)

What Changed?

● Added voice recognition:



Press the speech capture key (default: Windows Key + A) to start the capture. You will see a box that says "Recording..." in the top-left corner of your screen. Speak a word or phrase or sentence into your microphone. Capture2Text will automatically recognize when you are done speaking and will display a box that says "Analyzing...". The speech recognition will take a couple of seconds. When the speech recognition is complete you will see a list of possible transcriptions to choose from. When you choose a transcription, it will be stored in the clipboard by default.

When the results windows is displayed, you can press Enter to select the first transcription or use the number keys (1-9) to select the corresponding transcription.

To cancel a voice capture, press Esc.

To remap a key, edit settings.ini in the Capture2Text folder.

To toggle between 2 languages, you can either right-click the Capture2Text tray icon and select "Settings -> Toggle Voice Language" or use the toggle key which is Windows Key + Tilde by default. The languages to toggle between can be set in settings.ini.

● Added Send to Cursor Option:
This allows one to send the text to whichever textbox currently contains the blinking
cursor/I-beam. To enable, right-click the Capture2Text tray icon and select "Settings -> Send to Cursor".

● Minor Things:
- Renamed OCRAdjustment to OCRSpecific in settings.ini
- Moved the CaptureBox section in settings.ini to the OCRSpecific section
- Added VoiceSpecific section to settings.ini.
- Added StartVoiceCapture to Hotkey section in settings.ini
- Added VoiceLanguageToggleKey to Hotkey section in settings.ini
- Removed scaleFilter from settings.ini
- Removde the scaleFactor option from the menu (it's still in settings.ini)

Next Version:
I just noticed that there is a new version of the Tesseract OCR tool (v3.01) with improved vertical text support. After I test it, I'll add it to v1.08.

cb4960

Can anything in this long thread be of interest to a Mac user? I seemed to see .exe popping up from the very be beginning. If not are there alternatives?

jmignot Wrote:Can anything in this long thread be of interest to a Mac user? I seemed to see .exe popping up from the very be beginning. If not are there alternatives?

You might try the script bombpersons wrote for Linux users.

I have just posted version 1.08 of Capture2Text.

Download Capture2Text v1.08 via MediaFire (source code is included)

(Deposit Files Mirror)

What Changed?

● Upgraded the Tesseract OCR tool to version 3.01 which has better vertical text support and doesn't ignore small captures as much.

When the Tesseract versions of Chinese or Japanese is selected, you should specify the text direction (vertical or horizontal) using the text direction key: Windows Key + W. The text direction will not have any effect on the NHocr Chinese and Japanese dictionaries.

Before the vertical text support, the following capture would return 「し自な分よで努力」:



Now, when the text direction is set to vertical, it is correctly OCR'd from right-to-left, top-to-bottom: 「自分で努力しなよ」.

● Changed menu text for Chinese and Japanese to reflect the OCR engine being used.

cb4960

Has a thread been created where people talk about how they use this program? I'm interested in how people are using the voice recognition addition (I assume it's for furigana). Sounds like one of those sought after modifications for Anki, so instead of writing in the answer, just speak in the answer.

*note, I've pretty much been doing nothing study wise for 11 months so it's possible I'm way behind on all the efficient study tools development.

This looks awesome, but unfortunately I'm on a Macbook...
Let me know if somebody gets it to work on OSX

Nukemarine Wrote:Has a thread been created where people talk about how they use this program? I'm interested in how people are using the voice recognition addition (I assume it's for furigana). Sounds like one of those sought after modifications for Anki, so instead of writing in the answer, just speak in the answer.

*note, I've pretty much been doing nothing study wise for 11 months so it's possible I'm way behind on all the efficient study tools development.

To my knowledge no such thread exists. To be honest, I didn't have a particular application in mind when adding the voice recognition feature. I just though it might be fun to have on my PC after playing with it on my recently purchased Android device. Your suggestions are a good start though.

Zorlee Wrote:This looks awesome, but unfortunately I'm on a Macbook...
Let me know if somebody gets it to work on OSX

Unfortunately, I wrote this in AutoHotKey which is not exactly the best choice for portability.

I have just posted version 1.09 of Capture2Text.

Download Capture2Text v1.09 via MediaFire (source code is included)

(Deposit Files Mirror)

What Changed?

● Speech recording now gives you more time to start talking and should no longer cut off mid-sentence (unless you make long pauses while talking).

● The VoiceMaxResults setting is now properly implemented. Also, I changed the default from 5 to 9.

cb4960

http://img841.imageshack.us/img841/1117/tempqh.png

As a test, I tried three times to scan this character and got the following results:

.■ ■■ ■.
エ
門

LucidFaia Wrote:http://img841.imageshack.us/img841/1117/tempqh.png

As a test, I tried three times to scan this character and got the following results:

.■ ■■ ■.
エ
門

The Tesseract and NHocr OCR engines used by Capture2Text seem to have a higher success rate on more normal looking, smaller, less bold text, such as the text typically found in a manga or light novel scan (scroll 6 posts up for an example). I use this utility a lot with manga and have had a pretty good success rate. If you are still interested in using the software for such purposes, you might want to try a few more test points before giving up.

Using my copy on the character provided, Japanese (NHocr) consistently returns エ and Japanese (Tesseract) consistently returns a space character. Japanese (Tesseract) seems to ignore single character captures sometimes. It can help to capture two or more characters at a time to remedy this. Interestingly, Chinese (Tesseract) consistency and correctly returns 王.

(I wonder what happened to achieve 「.■ ■■ ■.」 )

Good Luck!

Hi,

This tool is working great for me!

I was wondering in the future if you would be able put an option in to make the Nhocr engine read top to bottom, left to right, just like the Tesseract one? When I'm reading manga, capturing multiple columns works with Tesseract, but I find the accuracy is quite poor. I hardly have to fix any characters with Nhocr, but then I am doing 3-4x the cut and pasting....

Thanks!

kisuke1 Wrote:Hi,

This tool is working great for me!

I was wondering in the future if you would be able put an option in to make the Nhocr engine read top to bottom, left to right, just like the Tesseract one? When I'm reading manga, capturing multiple columns works with Tesseract, but I find the accuracy is quite poor. I hardly have to fix any characters with Nhocr, but then I am doing 3-4x the cut and pasting....

Thanks!

I would like to have this feature as well. Unfortunately, NHocr doesn't officially support vertical text at the moment. Hopefully the author will return to the project in the future and implement it.

You're right that Tesseract isn't quite as accurate. However, the developers have an updated version of the Japanese dictionary file available in their repository:

http://tesseract-ocr.googlecode.com/svn/...raineddata (direct link, 24.6 MB).

Just download it and copy to Utils\tesseract\tessdata (overwriting the existing jpn.traineddata file).

I could be my imagination, but it seems much more accurate. I'm going to include it in the next release of Capture2Text.

Uhm, hello, I'm new here and I registered pretty much to ask your help!^///^"
I tried to run Capture2Text on my pc, but as long as I click on the .exe file I get this:
http://img202.imageshack.us/img202/2603/problemeoe.jpg

Do you have an idea of what the problem could be?

Thanks in advance!

Stardust Wrote:Uhm, hello, I'm new here and I registered pretty much to ask your help!^///^"
I tried to run Capture2Text on my pc, but as long as I click on the .exe file I get this:
http://img202.imageshack.us/img202/2603/problemeoe.jpg

Do you have an idea of what the problem could be?

Thanks in advance!

I think you've just uncovered an interesting bug.

Your profile says that you are from Rome, so have probably have an Italian keyboard, right? According to wikipedia, the italian keyboard doesn't have a backtick (`) character like US keyboards do. However, the voice toggle is currently mapped to the backtick character, so when the program loads, it doesn't know what to do with this character and behaves oddly from that point on. I was able to reproduce this problem by changing VoiceLanguageToggleKey in settings.ini to ò (which isn't on a US keyboard).

Solution:

1) Open up settings.ini (it's in the same folder as Capture2Text.exe) and scroll to the very last line which should look like this:

VoiceLanguageToggleKey=#`

2) Change this line to

VoiceLanguageToggleKey=#4

(notice that I replaced the ` with a 4)

3) Save settings.ini and try running Capture2Text.exe again.

Let me know if that doesn't fix your problem. I'll see if I can correct this bug in the next version (maybe I'll just use a non-backtick default).

Oh, thank you so much for your kind and quick reply!^___^ It works perfectly fine now! Thanks a lot again!

I have just posted version 1.10 of Capture2Text.

Download Capture2Text v1.10 via MediaFire (source code is included)

(Deposit Files Mirror)

What Changed?

● Added a real-time preview (shown in the top-left corner of the screen) of the actual text that will be OCR'd under the capture box. This eliminates guesswork and allows you to make adjustments (such as selecting the other Japanese OCR engine) before making the final capture.

If would prefer to disable the preview (or change its color, font, etc.), edit the appropriate settings in settings.ini.

cb4960

Hi there, cb4960. I have to say, great work on this wonderful tool. I use it all the time when I do translating work on some of the manga I work on. Unfortunately, my kanji vocabulary is not that great and when I use it for some of the handwritten kanji, the miss rate is pretty high. Would it be possible to implement a feature that recognizes handwritten kanji? If not, then do you know of any other OCR tool that I can use for something like this? Thanks in advance and much appreciated on all your hard work.

Nishimura Wrote:Hi there, cb4960. I have to say, great work on this wonderful tool. I use it all the time when I do translating work on some of the manga I work on. Unfortunately, my kanji vocabulary is not that great and when I use it for some of the handwritten kanji, the miss rate is pretty high. Would it be possible to implement a feature that recognizes handwritten kanji? If not, then do you know of any other OCR tool that I can use for something like this? Thanks in advance and much appreciated on all your hard work.

Capture2Text relies on two open source OCR engines: NHocr and Tesseract. Unfortunately, neither are very good at handwritten text at the moment. The only thing that we can do is hope that the developers of those engines continue to make improvements and someday add better support for handwritten text.

Hello cb4960, I am having trouble using the OCR as it seems every time I try to scan something the text appears backwards when using the Teserrac and not the NHocr.

Do you have any idea what i s causing this?

For the record this program amazing and love the work, now I need to test the Japanese Speech recognition

cosmochemistry Wrote:Hello cb4960, I am having trouble using the OCR as it seems every time I try to scan something the text appears backwards when using the Teserrac and not the NHocr.

Do you have any idea what i s causing this?

For the record this program amazing and love the work, now I need to test the Japanese Speech recognition

Tesseract has two text orientation modes: horizontal and vertical. Press Win-W to switch between them. If you are OCR'ing horizontal text when in vertical mode, the text will appear reversed as you describe.

Thanks for the awesome program cb4960! It took me a little bit of troubleshooting to get it working though.

I guess most of you already had some sort of Asian character set enabled in Windows, but I didn't have it enabled so everything just showed up as squares in the preview pane (and when I tried to paste to the clipboard). I had noticed a similar problem before when Asian characters in filenames would also be square blocks, but it didn't bother me too much so I never looked into it.

Anyway, I searched the internet for how to fix this problem and I followed the solution found here (quoted below the address):

http://answers.microsoft.com/en-us/windo...cadff4aafa

"So the trick is to change the system locale to Chinese (PRC) and then changing the system locale back to English (United States)."

That's what I did (actually I used Traditional Chinese but same idea), and it fixed the problem. Actually I didn't change it back to English yet, but I assume the Chinese characters will stay like the poster said.

Edit: By the way, I couldn't get the voice translate to work for some reason.. not sure why. I press control-A and it just gets stuck with the "Recording..." and doesn't do anything. The OCR is primarily what I was interested in, though.

JJ

I have just posted version 2.0 of Capture2Text.

Download Capture2Text v2.0 via SourceForge (source code is included)

What Changed?

●　Added the Preferences dialog. No more editing settings.ini by hand.

●　The popup window is now multi-lined.

●　Added option to preserve newline characters.

●　Limited preview to 150 characters. A trailing "..." will appear if necassary.

●　Added Speech Recognition Language option to right-click menu.

●　Cleaned up the right-click menu.

●　On the first run, inform user how to access the Preferences dialog.

cb4960

I was recommending this program to someone, but a question popped into my head: Does Capture2Text work on emulated games? I assume it does, but I can't test it out just yet. Things such as low resolution text in addition to wonky anti-aliasing results will make it harder I'm sure.

Anyone?

THANK YOU SO MUCH cb4960!!!
I was looking for a kanji OCR program for months now. Now that I have one now scanlating manga would be much faster.

I never knew such a program existed! O how long have I been tormenting myself trying to find those rare kanji in manga using all kinds of radical searches, once in a while thinking 'man would it be nice if I could just capture those characters directly, but that's impossible'. Well I guess it IS possible. Because You're a GENIUS. Thank you so much cb4960!