© 2024 Texas Public Radio
Real. Reliable. Texas Public Radio.
Play Live Radio
Next Up:
0:00
0:00
0:00 0:00
Available On Air Stations
KTPR 89.9 - Stanton will be off-air for a short time today for broadcast tower inspection.

Got A Voice For Radio? The Algorithm Speaks

iStockphoto

Nearly a thousand of you heeded our call on All Tech Considered to submit a voice sample. The idea: Let a computer algorithm decide if you have a voice for radio.

Now, we've got the results.

Actor Wilbur Fitzgerald rated highly (surprise, surprise):

But most of you who responded are not actors. And it turns out, you don't need professional training to impress man or machine.

Back in March, NPR reporter Aarti Shahani looked at a company called Jobaline. It creates software to assess the voices of job candidates for potential employers. CEO Luis Salazar says it's one additional metric along with things like experience and education.

For instance, a call center dealing with angry customers wants employees with nice, soothing voices. Retailers, on the other hand, want energetic voices to push merchandise.

We wondered how well the software works. So Salazar gave us a custom analysis looking for both characteristics.

He tells NPR: "This was a very interesting exercise. You want somebody who's engaging because it's a radio show, but you want something soothing, which seems to be the key characteristic of hosts of NPR."

Thanks, we'll take that.

And here are a few of the audience voices the computer algorithm ranked the highest. See if you agree.

Now, we have to admit, we were also struck by the voices that the computer did notrank as highly. According to our human ears, some of these low-scoring voices sounded pretty darn good. Take this one:

Salazar says he actually likes that one too, though he's not sure it's very soothing. "It's maybe the energy level," he says. "I think it would be a great voice for reading [an] audio book."

Bruce Raby, what do you think? A future in audio-book narration?

Ultimately, Salazar says there still has to be a human element here. Real people also judge these outcomes, and help the software get better.

Our thanks to all who put their best voice forward.

Copyright 2020 NPR. To see more, visit https://www.npr.org.