Better voices for the BART trains in the Bay Area!

BART hasn't had a change of personality since the turn of the century.

If you lived, worked, or been a visitor to the San Francisco Bay Area, you’ve probably taken the train at least once or twice. And you’ve probably heard George and Gracie as a result. They’re the robotic sounding voices you’ll hear on the platform, making train arrival and departure announcements.

At Rime, we’ve been daydreaming about the possibilities for integrating a modern text-to-speech tech stack into transit systems. So we thought: what if every time you took the train, the voices you heard over the public address system were ones that you had never heard before but sounded like they could be one of your friends or family members from the city you live in?

Meet Steven!

And Faith!

And Andre!

None of these voices belong to a real person but they still sound like everyday people you might interact with on the train you’re about to board!

---

The current voices of BART, Gracie and George, were cooked up by Lucent Technologies, a division of Bell Labs, around the turn of the millennium. Per a news story called Meet George and Gracie, the (synthesized) voices of BART, which incidentally was published exactly thirteen years ago:

Up until the late 1990s, the public-address system in stations was BART’s only means of communicating status information with passengers on platforms. The public-address system, with human-voiced announcements, was used for major events such as a train out of service – it didn’t have the capacity to do arrival announcements for every train at every station, even though that kind of information is very convenient for riders…

That’s where George and Gracie come in. BART chose a text-to-speech (TTS) system from Lucent Technologies, based on the long history of Lucent’s Bell Labs Division in developing TTS products. Lucent called its male voice John and its female voice Grace; at BART, they came to be called George and Gracie. The announcements alternate between the male and female voices on odd- and even-numbered platforms. The voices can be modified for pitch, speed, breathiness and other factors, and were tested with passengers on platforms to choose the versions that were the clearest and most pleasing to hear.

With all the powerful cultural attention paid to ‘generative artificial intelligence’ nowadays, it’s pretty easy to forget that generative artificial intelligence has been solving for these kinds of problems (and many others!) for decades.

Still, I think we’d all agree that Gracie and George sound, well, more than a little tired (though I’m sure very nostalgia-inducing for many people)!

Here’s another Rime voice:

Okay, just kidding, we never want to hear that last one again!

The BART announcements may be nostalgic, but there’s definitely room for an improved rider experience!