• works with colour video
• threshold slider added
• noise detection improved

1. Introduction
2. What do you need
3. Capturing subs from an avi (step by step tutorial)
4. Correcting mistackes

1. Introduction

Have you ever encountered a need to capture subtitles from an avi file?

You came across a copy of a movie in a language you don't master (and most of the languages we don't J ) and the subtitles are not available. You have two options: to find someone to translate it for you; or to find and use an already existing translation. We will talk here about the second option.

Most of the subs available came from the dvd source. But a lot of movies came out only on VHS tape or are just broadcasted on some tv channel. If they are ripped or captured we get an avi with hardburned subtitles. They are there, already well translated (hopefully) but is there any way to get them into some well known format (sub, srt) without too much manual work? This question bothered me for a while and after a looong time of searching, learning c++ and some clumsy coding I've developed a Virtualdub filter that does just that; it captures subs into idx/vob file (subpictures and time codes). This vob file has to be ripped with SubRip and we get a subtitle.

On this place I would like to thank Avery Lee for his excellent applicaton - VirtualDub and to Alain Vielle for his vob creating routines I took from his SubLog filter. Thank you, guys.

2. What do you need

• VirtualDub (I use the version 1.5.4)
• PodPis filter
• SubRip
• VobSub

• Subtitles Translator
• your favourite editor with spelling checker

1. Capturing subs from an avi (step by step tutorial)

First you install VirtualDub. Then copy PodPis.vdf into virtualdub/plugins folder where all external filters are kept. Open avi file with VirtualDub.

Crop the picture to the subtitles part only (apply Null filter for cropping)

To add a filter choose from menu Video --> Filters. An empty frame appears where the filters are added.

Click Add button and select Null transform filter. This filter does nothing, the only use of it is to crop the picture. Click Cropping button and increase Y1 offset so long that you get a picture slightly higher that a subtitle area.

Click Add again and select the PodPis filter. The configuration screen opens.



You only have to specify the top and the bottom of the two subtitle lines. First put a check to Show hint lines and than click Show preview button. The preview screen opens and all four red lines are on top with values 0. Play a bit with numbers and mount the bottom line to the bottom of the subtitle and the top line to the top of it. Here you can se an example.



Uncheck Show hint lines now and you should see a white subtitle on black background.

 

By increasing or decreasing the Threshold slider you emphasise the letters. Decrease the threshold value when letters gets disconnected (see "m" on the picture) as disconnected letters are not recognised properly by SubRip. Increase the threshold value if there is too many background noise among the letters.



 

Close filters and you are ready to start with capture. Just click Output playback (the third bottom button from the left, the one with a small triangle and a small 0) and the detected subs are captured into PodPis.sub and PodPis.idx files. They are stored in c:\ folder (I have to make that the user could select the output file, next version J )



Go to sleep. This step takes the same time as the avi lenght (or even slightly more).

After the whole avi has been processed we have to rip the .idx/.sub file with SubRip (.idx/.sub is the same format of the subtitles as on DVD). But due to some unknown bug in Alain Vielle routine, the .idx/.sub files are not completely readable with SubRip but this can easily be corrected with SubResync (part of VobSub package). Just open Podpis.idx in SubResync and save it with another_name. Close SubResync, we don't need it anymore.

Now open this another_name.idx file with SubRip and start ripping. The picture in avi file often flickers and that causes a slight difference between letters of the same kind. You will have to type "a" for an "a" more than once. I suggest that you change the Sensibility on Character Difference to 970 (Options--> Advance OCR Setup). This takes the most of the time.

2. Correcting mistakes

For suggestions and questions write to: PodPis at hotmail.com (replace at with @, try to trick spammers off)