Radio Drops

Radio Drops #6: A.I. Steve

By

Steve Warren

·

·

3 min watch · 3 min read

In a dark studio, two identical grey-haired men in their seventies face each other like a mirror image, each leaning in toward one shared vintage microphone on a stand between them; both are pure flat silhouettes against a lit band behind them, the heads solid single shapes with no features at all, the right-hand figure slightly translucent like a copy; background is a dark wall and one soft lit band; props: the microphone; medium shot.

We Fed My Voice to A.I. and A.I. Came Back with…Me

A pal fed my voice into an A.I. voice generator and had it clone me. He gave it some promo copy and the A.I. voice popped out another “me” saying the words. Hear a couple of examples here. The video runs 2:59.

Watch the video on YouTube: Radio Drops #6: A.I. Steve

Real Me or A.I. Me?

My thinking on A.I. voices is changing on an almost daily basis. I’ve been saying that people can always tell it’s an A.I. voice, and up to this point they can. People do not like listening to A.I. voices; we’ve already seen research on that.

So the big broadcasters and the big agencies that are using A.I. voices to replace people now will eventually learn that’s penny-wise and pound-foolish. If you’re going to sell something, you need a real human selling it. An A.I. can’t sell very well, because an A.I. just sounds insincere.

But that’s changing. I just heard ChatGPT’s new A.I., and she’s almost as good as Scarlett Johansson’s A.I. in the movie Her, which was as human as can be.

And there’s also this. From time to time I do some voice tracks for a pal who’s got a hot country radio station. He likes to have stagers introducing the new acts he’s added recently. He calls them his New Crop Country singers. He fed some of my voice tracks into the A.I. machine, and it came out with this. Listen and tell me: is it real, or is it not real?

I know a guy in Florida who has a radio station on the Gulf Coast, and he does a daily shift on his station, a four-hour shift, I think. A couple of weeks ago he got a cold, a sinus infection and a sore throat, and he could not do his daily shift.

Now, he’d been playing around with ElevenLabs. Remember that if you want to play with it: ElevenLabs, with the eleven spelled out. It’s a voice generator. He had been training ElevenLabs with his own voice.

So he typed up his ad-libs and the things he was going to say between the records, fed the copy to ElevenLabs, and ElevenLabs gave him his voice tracks. He put them on the air at his radio station, and now his personal A.I. voice is doing his air shift for him when he can’t.

It’s a new world. A whole new world.

SW

Written by

Steve Warren

Steve Warren has programmed radio since 1964. He built MdB, the first music scheduler for the Mac, in 1987 and Music 1 for Windows in 1994, and is a founding partner of MusicOne.

Put it to work on your own station.

Bring your library to MusicOne Cloud and schedule a real day.

Free for 30 days. No credit card, nothing to cancel.