Subresync new feature ocrdll.dll

[kxy]

The new feature automated OCRing process sounds cool. But unfortunatly it didn't work. It seems to have trouble with </i> characters so far, might have problem with others...

I am not sure if this is SimpleOCR's problem or not. If so, can we manually edit this OCR library (dll, ENGLISH.WDC) to make it work?

Here is an example:

3
00:01:26,219 --> 00:01:31,213
f Man #2) Owl

f Womanlshut up! Keep still

while the doctors? examlhlhgyou.


This suppose to be

[Man #2] Ow!

[Woman] Shut Up! Keep Still
While the doctor is examing you.

[gabest]

Right, ocrdll is outputting rubbish in about 10% of the time :) I'm afraid dvd subpictures are not the ideal source material for OCR... Mainly the wrong detection of spaces messes up the dictionary lookup...

To make it somewhat better, in the future I'll try to utilize the recogized characters with subresync's manual OCR engine and construct a character matrix automatically, or at least a part of it.

[kxy]

Gabest, from the readme file,
- You can save the recognised characters into a file and reload them for editing or OCRing later?

So this isn't working yet, right? Can you make subresync import *.sum off SubRip, that would save us tons of work?

[gabest]

Do some manual OCR and then you can save *.vsocr files. Until the database is empty this last option won't show up

[kxy]

Wow, your reply is fast, I didn't expect that, well I edit the msg. Please read again. :P

[gabest]

Well, last time I looked subrip's delphi code I found that it had some kind of wiered file format, so I decided not to do the reader function...

[kxy]

So import *.sum won't happen then? I should start building my own vsocr database?

Another qustion, is it possible to do italic with Subresync? By using </i>?

[gabest]

I think there is no reason to keep vsorc files to long around, I've never seen two identical subtitling.

And no, no italics yet.

[kxy]

The reason to keep it, is if you find some mistakes, you can always go back and edit it. I have a big *.sum file, using subrip, it is able to recongize most of the characters, out of the entire movie, only a few char need to be OCR.

I will give another example, before you have the vsocr work around. I would wait till I have 20+ *.sub then do the OCR and leave subresync open and sure it speeds up the process quite nicely, cause it is able to recongize most characters.


I noticed when open up *.vsocr it is sorted by alphbatical order, can you make it sort by the "order of entering"? That way it makes editing easier.