Doom9's Forum - View Single Post

Saligia · 12th September 2005, 06:05

Hi!
First of all, thanx for devoting all these hours, weeks, months... keeping Subrip going strong.
It is an one-and-only. And without it we would be nowhere.

THX!

Lately I've come across some subs that, no matter what, just would'nt be OCR'ed to anything, but endless lines of ocr-errors.

Born out of pure frustration, I came up with this solution, that works just fine.

1. Rip subs to BMP/SON with Subrip.
2. Convert to SUB/IDX with Son2vobsub.
(Can be scaled, to obtain large, clear, non-joined chars.)
3. Load SUB/IDX into AVIsynth on blank clip, with movie length.
(White chars on black background)
4. Make fake AVI with MakeAVIS.
5. OCR the .avi with Subrip.

It works well, thoug it's no timesaver. But then again, I'd rather spend some hours clicking the mouse now and then, than powerbashing the keyboard until i breaks and burn.

And now to my real agenda. (Of course there had to be a catch!)

When OCR'ing "hardsubbed" txt this way, it seems that Best Guess is correct, next to always.

So these options would be Soooo cool:
A. Use Best guess after (n) seconds, if not interupted. 0=allways
B. Insert (selectable chars) if best guess=none
C. Comment in lines where best guess is used

This could even make the OCR'ing unattended, and easy to correct afterwards with some spellcheking.

Thanks for your great work.

12th September 2005, 06:05	#199 \| Link
Saligia Registered User Join Date: Jun 2005 Location: Denmark Posts: 11	Wish it was there... Hi! First of all, thanx for devoting all these hours, weeks, months... keeping Subrip going strong. It is an one-and-only. And without it we would be nowhere. THX! Lately I've come across some subs that, no matter what, just would'nt be OCR'ed to anything, but endless lines of ocr-errors. Born out of pure frustration, I came up with this solution, that works just fine. 1. Rip subs to BMP/SON with Subrip. 2. Convert to SUB/IDX with Son2vobsub. (Can be scaled, to obtain large, clear, non-joined chars.) 3. Load SUB/IDX into AVIsynth on blank clip, with movie length. (White chars on black background) 4. Make fake AVI with MakeAVIS. 5. OCR the .avi with Subrip. It works well, thoug it's no timesaver. But then again, I'd rather spend some hours clicking the mouse now and then, than powerbashing the keyboard until i breaks and burn. And now to my real agenda. (Of course there had to be a catch!) When OCR'ing "hardsubbed" txt this way, it seems that Best Guess is correct, next to always. So these options would be Soooo cool: A. Use Best guess after (n) seconds, if not interupted. 0=allways B. Insert (selectable chars) if best guess=none C. Comment in lines where best guess is used This could even make the OCR'ing unattended, and easy to correct afterwards with some spellcheking. Thanks for your great work. Last edited by Saligia; 12th September 2005 at 06:23.