Most speech-to-text programs don’t work that well. That’s like actually a big secret. I worked in transcription for a long time. I tried to get this stuff to work many a time. I knew doctors who, out of frustration after spending a bunch of money for speech-to-text software, went back to using a transcription service.
The speech-to-text services that seem to work are SERVICES and guess what? They use human transcriptionists. A lot of the time it’s an overseas service. They’re listening to whatever it is and transcribing it. A lot of hospitals, clinics, lawyers, and cop departments are using this type of thing now. And sometimes the transcription is VERY, VERY BAD. A lot of the time it is. Remember Rush Limbaugh talking about weird stuff that was coming up on his voice recognition thing that he was using in the car? I sent him an email about that. Someone from WND ended up contacting me. (I don’t know what happened after that.)
Transcriptionists like I used to be get stuck editing this stuff for 10 cents on the dollar. You edit stuff that’s insultingly bad and you know it would have only taken you 5 minutes to transcribe it if they just had let you do it in the first place.
Gov Romney was heavily invested into crap like this, ie, overseas teams of “health care workers” being cheaper and easier to hire rather than American workers. Transcription is one of the ways they got that whole thing started, because understanding what the doctors are saying is of course very important, AND you can easily work from a remote location. That’s why they called it “Romneycare” instead of “0bamacare” at first.
They’ll never tell you that your voice commands and speech-to-text are actually people. It’s a process like captioning movies and so on. They want you to think it’s the AI.
So you’ll go out and waste money on the AI, trying to save yourself some time and hassle.
I’ve known a few people who could get Dragon to work well or the “speech recognition” that comes with Windows, or whichever one you name that is all computer driven without the intervention of a human. But it takes a long time. It’s like teaching a really stupid little kid how to talk. A really, really stupid little kid.
bkmk
I’m a court reporter and my latest software program includes an AI Boost with Google talk to text.
What it does is anything I miss when I am taking down the transcript the Google will type it into the transcript. It works pretty well for attorneys and witnesses with good speaking voices. Attorneys and witnesses with smaller, less robust voices are not captured very well, if at all.