Creative Communities of the World Forums

The peer to peer support community for media production professionals.

Activity Forums Apple Final Cut Pro Legacy OCR subtitle text?

  • OCR subtitle text?

    Posted by Peter Pop on August 3, 2010 at 11:28 am

    Hi folks,

    I’ve just had an unusual job request from a client. He has 60ish hours of DV footage with burned-in subtitles (English) and wants to have the text transcribed and relayed over matching HD footage.

    He has no text or file version of these subtitles.

    Short of me hiring someone to manually transcribe all of this, does anyone know of any reliable video OCR method that could be used? I.E. is there any software that can ‘read’ subtitles and textify them?

    I know it’s a long shot… but you know… i’ve been surprised by technology before.

    Dennis Radeke replied 15 years, 9 months ago 4 Members · 7 Replies
  • 7 Replies
  • Chris Tompkins

    August 3, 2010 at 12:57 pm

    I googled and saw this:

    https://www.sri.com/esd/automation/video_recog.html

    Chris Tompkins
    Video Atlanta

  • Peter Pop

    August 3, 2010 at 1:11 pm

    I had a look through a bunch of the google results before i came here, and invariably they where all propriety, in development or custom built camera systems.

    I couldn’t find anything ‘buyable’ never mind anything that appeared to run on OSX or WIN.

    Still… i’ll keep digging.

    Cheers

  • Mark Suszko

    August 3, 2010 at 1:43 pm

    I’ll tell you how I might do it if it were me.

    I’d buy a copy of Dragon Naturally Speaking, train the software to my own voice, and just read the titles into the headset mic while playing live. Kind of brute-force, but should get the job done. You are consigned to transferring these clips anyhow, as you load them into your system; might as well do the transcribing at the same time.

    I’m not very familiar with the latest Adobe Premiere version, but I’m led to understand it has some new script-based editing capability that might help with this as well, in terms of synching the spoken-word to text conversion to the time code of the clips.

    Whatever YouTube is using to create their automatic captioning, it is very good.

  • Peter Pop

    August 3, 2010 at 2:00 pm

    Thanks for the idea.. being Scottish I have worries about voice recognition.

    This You-Tube vid might help explain why 😀

    Another idea I was given was to screen grab each subtitle (the client already has then digitised in .dv) to jpg and try a batch OCR on them.

    Thanks for the head up on the premiere thing… I’ve got CS5 arriving later this week… so maybe my answer lies therein.

  • Mark Suszko

    August 3, 2010 at 2:28 pm

    Och aye, Funny. The trick with voice recognition is that it works best trained to just one voice. Just because of the higher frequency range, I think ’twill ken a bonnie lass more clearly than a man’s brogue.

  • Peter Pop

    August 3, 2010 at 2:46 pm

    I think I may have found a solution.

    AviSubDetector.

    It looks a wee bit rinky-dink but if it saves me £££ in transcribing then I’ll work through it.

    I’ll update this if i have any success.

  • Dennis Radeke

    August 4, 2010 at 1:39 pm

    Yes, Premiere Pro CS5 has the ability to input your own text to align to a given clip, or if a script is part of the production, to align the clips to the script. This is in addition to the speech-to-text functionality that was provided in CS4.

We use anonymous cookies to give you the best experience we can.
Our Privacy policy | GDPR Policy