1 Reply Latest reply: Aug 5, 2009 2:09 PM by _durin_ RSS

    Transcribing only works for short clips?

    bspence11 Community Member

      I tried transcribing about an hour's worth of audio and it seemed to work, but it didn't appear in my audio files' metadata. I was able to do it for a ten minute chunk, but that's it. Is there a more robust way of getting my transcribing work done? I was really happy to hear that Soundbooth had the ability to transcribe audio, but if it can't do large amounts of audio, I'll have to look for another solution. Can anyone get this feature to work with a lot of audio? I only made an attempt with one hour's worth of audio. I actually have meetings that last for eight hours or more that I'd like to have transcribed.

       

      Also, does the transcription accuracy improve over time if I correct the transcription afterwords? It seems pretty poor at recognizing some words (although, admittedly, my audio source is rather poor).

       

      Thanks so much for your help!

        • 1. Re: Transcribing only works for short clips?
          _durin_ Employee Hosts

          Did you save the file again after performing the transcription?  I don't believe it will automatically be saved with the file after the transcription process is complete.  There's no time limitation, although with very large files (upwards of 4 gigabytes) the addition of metadata may force the filesize too large and that might prevent it.  This doesn't sound like the problem you're seeing, so I would suspect it just wasn't saved.  The 8-hour meeting, I might suggest breaking into smaller pieces anyways, because you are very likely to run into the file size ceiling if/when the file is conformed.

           

          As for the accuracy improving, at this time the dictionary files are static and do not implement any learning behavior.  Obviously, the higher quality audio source, the more likely you are to get more accurate transcriptions, but in general, this tool should be used more as a guide and less as an accurate transcript of the audio.