AutoHotkey Community

[Mkbailey755]

My girlfriend works for special needs kids one boy in her class is a big fan of a cartoon called naruto. the problem is he cannot read what so ever and the newer episodes are sub titled(Japaneses) so the idea came to me to take an OCR to read sub titles and import this into speech. so the real ? is this a reality or is OCR much to slow? how hard of a task is this?
thanks in advance
sorry if this is in the wrong section i didn't know if this should go in general chat

[Z Gecko]

I am no OCR expert, but from my knowlegde i have to say:
It´s probably not possible.
You would need a combination of high speed OCR (for scanning 25 to 60 pictures per second) and
a very good engine to distinguish the text from the permanetly moving background.
If someone really made something capable of this, it´s probably comercial and expensive. For shure not possible with pure AHK commands.

But you got one chance: Movies often have seperate Subtitle files(like the dvd), i don´t know if they are plain-text or kind of a video-file. You could maybe extract them(if txt-file) or have an OCR-programm scan them, in both cases probably befor you watch the video, not meanwhile.

[SomeGuy]

yeah extracting the subtitles would definitely be what i would look at.

[[VxE]]

Hardsubs (the subtitles are directly edited into each frame) would be difficult for an OCR to catch, given that the video might be displayed with help directly from the video card and hence an OCR program might not see it.

Ideally, you could get the softsubs (an extra text file with the subtitles along with timing info), then use AHK to parse the softsub file and send the subtitle strings to Window's text-to-speech thingy at the proper times.

I believe some media players have an open API through which an AHK script can obtain timing information (i.e. SendMessage to get the current frame number) to sync the subs-to-speech with the video.
_________________
My Home Thread
More Common Answers: [1]. It's in the FAQ [2]. Ternary ( a ? b : c ) guide [3]. Post code inside [code][/code] tags !

[IsNull]

Other suggestion:
If you can't extract the subtitles, you can try the following:

You make a screenshot from every frame. (I'v got written a wrapper for "ffmpeg", so you can use this to generate screenshots) Then, you only have to analyse the pictures. And this ist supportet from all OCR Tools.

regards
_________________
http://securityvision.ch

[Oberon]

Go to another fansub group for .srt's and Loop, Read them. Dattebayo can't be the only one, surely...

[Mkbailey755]

Thanks man I had to do a alittle research on srt's but I think that would work Im at work right now but I fool around with it this weekend