It was one year ago this week that I returned to the radio after a two year absence, with a new computer generated voice which we call 'Jamie Dupree 2.0' - a high tech invention which has allowed me to continue my radio news work, even after an unknown medical problem took away my voice.
After doing your job one way for over thirty years, it has taken a little time for me to get used to operating the controls of Jamie Dupree 2.0, which produces a voice that sounds like me, but no matter how advanced it is, it can also go haywire in a split second and sound like a bad robot.
1. First, a recap for those who might not know the story. In the Spring of 2016, I was covering the race for President. Everything was fine. I took an Easter vacation with my family, got a stomach bug, and suddenly my voice started having problems. It's now to the point where I can barely talk, with the diagnosis being a neurological dystonia - the signals from my brain to my tongue and throat are getting messed up somewhere along the way, and my mouth just won't work correctly to form words and sounds associated with speech most of the time. My company found a firm in Scotland, CereProc, which built a computer version of my voice from my audio archives, using tapes from my old radio news stories.
2. Jamie Dupree 2.0 just isn't just typing some words.
It would be nice if I could just type my radio scripts, hit a button, and magically have a perfect audio file for my next radio newscast. But it's a bit more complicated than that, as the field of computer generated voices is still in its infancy. Luckily, there are special computer commands which can be employed with the Text-to-Speech program that runs with my special voice. Those commands allow you to slow words at the end of sentences, mimic the more natural ways that we speed up and slow down during regular speech, and find ways to make the overall sound less robotic. It makes for some clutter on screen, but this is what one of my typical radio news scripts might look like:
3. Some Jamie Dupree 2.0 words just don't sound right. For whatever reasons, there are some words and phrases which don't come out right when you type them in. The last name of Rudy Giuliani works perfectly, but the last name of Commerce Secretary Wilbur Ross does not - I have to use "Ros" instead. Robert Mueller's last name came out as "Myoo-ler," so I had to spell it as 'Muller' to get it to sound right. Those are just a couple of examples of how, over the last year, I've had to do a lot of experimenting to figure out how to sound out certain things, which can be frustrating when you need to get a story done for the next newscast. Like one of the Democratic hopefuls for President, Indiana Mayor Pete Buttigieg. That's hard for most people to say, let alone for me to try to figure out how to get my computer voice to say it. So far, I've settled on this: Pete Budda jidge. Here are some other names and words that I have to get creative with:
4. Stumbling on ways to make the 2.0 sound more real. As I did more and more work over the last year with my new voice, one of the most obvious problems was trying to figure out how to mimic our own real voices, as we speed up and slow down certain words and sounds. XML commands dealing with 'emphasis' didn't really work. I tried using my audio editor program to add some emphasis to my words, but that only was a minimal success. By accident, I found that by repeating a word or phrase, it would sound different - sometimes I would get exactly the right feel and emphasis.
In the following example, I wanted a little more 'oomph' for the phrase, "President Trump tweeted from Air Force One." When you watch the video below, you will hear the original version of the audio produced by Jamie Dupree 2.0, followed by the version where I had it say the same phrase five times in a row. Then the two are compared at then end of the 30 second video. For some reason, the repetition creates a little more emphasis. Why? I don't know. I don't really care. All I know is that I found a shortcut which makes it sound better.
5. The reaction to 2.0 continues to be mostly positive.
Remember, these are the days of social media, so it's not difficult to make your voice heard, and tell me that you never liked me in the first place, and you're happy that my real voice doesn't work. But those messages only spur me to keep going and to work harder at being heard on the radio. I don't want to be using this computer voice technology, but thankfully for my family, it's available, and it has allowed me to continue in my career as a radio reporter covering Capitol Hill. As I have detailed above, it's not a simple process to get a story on the air. It takes time to mold the words into the correct sounds, and get that into our radio newscasts. It would be much quicker to just open my mouth, hit the record button, and start talking. For whatever reason, my brain won't allow that to happen.
Support real journalism. Support local journalism. Subscribe to The Atlanta Journal-Constitution today. See offers.
Your subscription to the Atlanta Journal-Constitution funds in-depth reporting and investigations that keep you informed. Thank you for supporting real journalism.