Tools used: SpokenMedia
From time to time I’m asked by Tufts clients if there is any way to automate the creation of a text transcript from an audio file. Recently the media in question was in the form of a DVD recording of a lecture.
My first port of call for a request like this is the SpokenMedia service at MIT. These folks have developed an application that attempt to do just that.
They ask for video files (e.g. avi, mp4, mov etc) but I’ve uploaded MP3 files also.
In this case I first had to have one of my colleagues help me rip the audio off of the DVD file (thinking about this now I should have just used a DVD ripping tool to migrate it into a mp4 video, oh well). Once I had the file I followed the directions on the SpokenMedia upload page.
Note: once you’ve uploaded the file you have to email mckinney (at) mit (dot) edu as the email confirmation component does not work as advertised.
Also note: The system is VERY sensitive to audio quality and the accent/enunciation of the speaker. Obviously, the better the audio and the more clearly the speaker speaks, the better the transcript.
Clients requesting this type of service tend to say that “something is better than nothing” so we go ahead anyway!