When StreamGuys built SGcreative this year, we were more concerned with whether AI should produce audio than whether it could. Until 2026, we had stayed on the sidelines of producing AI audio. The use case and quality needed to be right.
The right use case is the kind of audio a person should not have to sit and read on repeat. Weather and traffic are the clearest example. A live host reads them fine; the trouble is how fast the information turns over and how many places it must run at once. The forecast changes through the day. Traffic goes stale in minutes, and every market wants its own version. To keep all of that accurate everywhere, in English and Spanish, you need someone re-recording the same segments repeatedly, and the cost gets out of hand fast. That is the work synthetic voice is built for. SGcreative pulls in live data on weather, traffic and local conditions and turns it into sponsored audio that fits wherever the listener happens to be. Someone in Los Angeles gets their forecast; someone in Chicago gets theirs. The spot refreshes on its own instead of aging on a server.
Connection is where the answer changes. Our customers reach audiences across radio, streaming and podcasting, and people trust that programming because it sounds like them. You hear it in the rhythm, in the way a host moves between English and Spanish inside one sentence, in references that only land if you live them. That is why people tune in and connect. Hand those hosts to a model, and you would save some money while gutting the thing you were selling. So, we left them alone.
The idea is to let each kind of voice do what it is good for. The personalities hold the relationship with the audience. The automated layer takes the repetitive, hyperlocal material that changes by the minute, and it runs at a scale that makes national advertisers willing to buy into local, culturally specific audio. Publishers get into campaigns that used to pass them by for being too small or too scattered.
This is roughly where I think the business is going. Stations that reach for synthetic voice as a cheap way to shed talent will hollow out the reason people listen. The ones who use it to clear the grunt work away from their hosts, so their best people spend their airtime on what only they can do, will come out well ahead. The tools are good enough now that quality has stopped being the argument. What is left is judgment about where a synthetic voice belongs.
Radio has always run on voices people believe. The difference now is that we can decide which of those voices needs a human behind it.
The author is executive V.P. of global sales for StreamGuys.
This story originally appeared in the RedTech special publication Radio Futures: The Intelligent Signal. You can read or download that edition for free here.
You can access all past RedTech publications, also for free, here.
These stories might interest you
Radio Futures champions the intelligent signal
StreamGuys removes delay for Bats
Tech Focus: StreamGuys consulting service aids programmatic growth
