Voice Recognition Software to a CC encoder

More channels means more money for the techs, right?
Post Reply
w4cl
Posts: 477
Joined: Wed Apr 09, 2008 3:04 pm
Location: 1 3/4 meters

Voice Recognition Software to a CC encoder

Post by w4cl » Wed Jul 09, 2014 2:31 pm

Has anyone tried using a program like Dragon and figured out a way to send the output to a closed captioning encoder? We are looking at a way to do live captioning without paying someone to do it.

KE0MI
Posts: 31
Joined: Wed Aug 04, 2010 8:38 am

Re: Voice Recognition Software to a CC encoder

Post by KE0MI » Thu Jul 10, 2014 8:21 am

http://www.enco.com/products/encaption3.html?

I have been on-board with this product since the rules were established. The first generation stuff would actually poll the scripts from iNews and add words to the dictionary that it knew would be coming. It did at best 80-85% accuracy on my best anchor. (The person was male with a very clear speaking voice, in a controlled noise environment). There were times we did get calls demanding that I fire the captioning person, as well. I told them that that person would never work here again!

The generation 3 version does much better, but still not 100%. Any background noise or music will make the captioning questionable. If you really want a chuckle, read the dictionary it has on board of words (phrases) that are NEVER to be captioned. I have expanded my vocabulary immensely when I need to get someones' attention.

The problem with Dragon (or the variants) is that you still need a shadow talker - someone to train the software to their voice, then listen to what is being said, then repeat verbatum in a voice the software understands. An expensive approach and a new position at the station that would have to be present whenever you need captions.

I went to a hybrid mode. We use the scripts when provided, Encaption when no scripts are present, and a captioning service that listens to the program audio and live captions the feed when near 100% accuracy is required. A three way data switch allows the prompter operator or director to switch the data going to the encoder. I guess it could be controlled by the Ignite! software, but where is the fun in that?

I do not own stock in the company, nor do I speak on the behalf of my employer.

w4cl
Posts: 477
Joined: Wed Apr 09, 2008 3:04 pm
Location: 1 3/4 meters

Re: Voice Recognition Software to a CC encoder

Post by w4cl » Fri Jul 25, 2014 6:55 am

One of our stations uses Dragon with excellent results. We are trying to get exactly how they are using it. All I know, we have had them dial into our audio feed and they have CCed a couple of Sunday morning shows for us and the correction rate is over 97%. It will get tripped up on specific locale names (how would you pronounce "Pfafftown"? For the record it is POFF-town.), but other than that, it is really amazing.

I got a "quote" from Dragon and it was amusing how they wanted to sell us several servers and exotic software for around $100k and I know for a fact our station that is using it didn't spend more than $7k to make theirs work. And then the Dragon sells guy bugged the crap out of me for two weeks wanting to know if we were going to buy it. I email him back saying $100k was WAY out of line and did they have a cheaper way of doing it. No replay for that one.

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest