see discussion here
Not sure how well it would work though. I'd need to find out exactly what the speech recognition engine expects of its input stream.
Not sure how well it would work though. I'd need to find out exactly what the speech recognition engine expects of its input stream.