I've been working on transcribing a lecture series from audio to text files. There are just over 20 lectures, and each one is about an hour long. Using VLC Player, I'm able to slow down the speaker's words to a rate that I can keep up with.
However, it's extremely slow going, and it takes so much concentration that I can barely get through a third of a lecture before I have to take an extended break.
The thought had occurred to me that I would be better served by using some kind of speech recognition software, and then proofreading the computer-generated result.
This is likely to be a one-time thing, so while I don't mind using commercial software to get the job done, it would be preferable if it wasn't too expensive (or free; free is always good).
Command-line solutions are fine, too (:
Any suggestions?
Best Answer
My experience is that unless you get a very good speech to text solution, you are simply letting yourself in for a lot of frustration. You'll be spending more time making corrections to garbled transcriptions than you would have spent typing from the start.
There's a reason there are not many effective solutions on the market: this stuff is HARD!
The Dragon engine from Nuance is the best I've tried.