[nmglug] suggestions/help if I'm on the wrong track

[Sam Noble]

On Sun, Apr 08, 2012 at 10:54:59AM -0600, Steve Katona wrote:
> Here's what I want to do and wonder if there is software that will do it or
> if there is a better way than I have in mind? A friend has posted about 115
> YouTube talks that run from 5 minutes to near 30. A selection of them will
> make a book that we think will sell but more importantly would be important
> to be available. I transcribed one by starting and stopping the video and
> typing (at my usual 25 wpm speed). This is tedious and time consuming.
> Hiring someone to transcribe is a possibility.
> 
> My question is does anyone know of voice to text software that can listen
> to the YouTube being played and put that into a text file that I can
> manipulate with OO or LibreOffice? Is there a better way?

The only significant player I know of in free (as-in:) speech
recognition, is sphinx/pocketsphinx (http://cmusphinx.sourceforge.net/)

Unfortunately it's not especially usable.

This looks like the closest thing to what you're trying to do:
https://live.gnome.org/Gaupol/SpeechRecognition A nice gui subtitle
editor with built in pocketsphinx.

But I futzed with it for a while and haven't yet managed to get a
working version of the gstreamer plugins for pocketsphinx going.
Plus I have no idea how it well it would work if you did get it built or
found some working binaries. Sphinx is pretty good at reading back "GO
FORWARD 10 METERS" but I've never really seen it in action on live
conversational speech.