Xaviar ‘X’ Jernigan, the voice of Spotify’s DJ, explains what it’s like to become an AI

Image credit: @xavierjernigan

In March, Spotify launched its first AI-powered feature with the debut of AI DJ, a smart audio guide with compelling, lifelike audio. It turns out that the AI ​​persona was actually based on a real person. Xavier “X” Jernigan, Spotify’s Head of Cultural Partnerships, was honored to be the first voice model for the AI ​​feature.

TechCrunch interviewed Jernigan to learn more about the process of training AI and future plans for Spotify’s AI DJ efforts.

The new AI DJ personalizes the listener’s music listening experience and curates a selection of music based on their interests. It also gives commentary on each song, just like a real radio host.

In addition to his main role on Spotify, Jernigan is also the host of various Spotify podcasts, including “The Window,” “Showstopper,” and the now-defunct podcast “The Get Up.” As such, he is used to having his voice heard by millions of listeners. Still, having his voice remembered as his AI is a unique experience.

Spotify chose Jernigan as its first voice acting model because his “voice and personality already resonated with so many listeners,” Jernigan told TechCrunch. “[The company was] I’m pretty sure I would resonate this way too.

Spotify’s morning show “The Get Up” has nearly 6 million listeners and remains the top 10 podcast on Spotify until its abrupt termination in 2022, demonstrating Jernigan’s appeal.

Still, being a voice model for a DJ was hard to understand at first, the podcast host admits.

“I was offered a voice acting model for a DJ, and I was really surprised when they explained it to me,” said Jernigan. “The first time I heard this I had nothing to see and I was like, ‘Wait what? C’mon, and is it AI?”

“For me, working with AI this way was a new experience. I was just blown away,” he added.

Spotify says its AI DJ was built using both Sonantic and OpenAI technologies.

Sonantic is an AI startup acquired by Spotify last year. The company’s technology was responsible for creating realistic AI-based voices, including the one used for Val Kilmer’s voice in Top Gun: Maverick.

Prior to the acquisition, Spotify spent several years researching AI-powered technology and worked on DJ features “in several iterations,” Jernigan said. He declined to share the exact time the process took, but said integrating Sonantic technology “really sped it up.”

Jernigan explained the process of training the AI. This required going into the studio, reading the script, and conveying different emotions with different rhythms and inflections. He gave his AI certain words that he used only to make it feel as authentic as possible.

“I use the words I say…I don’t say ‘song’ to a song. That’s not my way of speaking,” he said. “I say ‘hit’ or ‘bangers.’ So you hear DJs say those words,” continued Jernigan. I went through the whole process, including how to say hello. I would carry around a notebook and write down different phrases I might say. “

He added that the Spotify team made sure the AI’s voice maintained his natural pauses and breaths so that it sounded truly human.

Even Jernigan’s mom gave the results her stamp of approval.

“[DJ] Passed the mother exam. I played it for her before it came out and explained to her, “She listened to all my podcasts, so she heard me record and play my voice before. So I said, ‘That sounds like you.’ My mother said it sounded like me, so I knew it was spot on.

Realistic AI voices already exist, but Spotify’s DJ is arguably the most soothing and soothing sound compared to anything else I’ve heard. Google’s Duplex technology may sound authentic, but it’s not always nice to hear when you’re trying to swell to your summer jam playlist.

“For me, acting as a voice actor, my goal was to connect with people, talk to people, and think about one person. I imagined talking to someone and being friends with them,” he added.

Not only is the AI’s voice easy to hear, but the design of the DJ itself is also friendly.

The animated green circle that users see when listening to a DJ nods to the Spotify logo and moves like a mouth when the AI ​​speaks.

“When it comes to design, we thought about the whole experience: how it works, how it sounds, how it looks, and how we make it personal to each user,” he told TechCrunch. Visually, we explored some options early on that felt more technical (think sound waves, etc.), but we wanted to humanize the AI, so this was right. I felt no…”

“We wanted it to look and feel unique. It was so unique, in fact, that we patented the design,” Galloway added.

Jernigan contributed to DJing in other ways besides recording his voice.

For AI to provide expert commentary on music, Spotify has put together a writer’s room of curators, cultural experts, and music experts.

Due to Jernigan’s extensive background in music, he was also participating in the Writers’ Room. He has previously worked with top his artists like Diddy, Amy Winehouse and 2 Chainz.

Jernigan is the DJ’s first voice model, but listeners may hear more voices in the future.

TechCrunch asked Jernigan if the company plans to adopt voice models that speak other languages.

“Look forward to it,” he hinted.

AI DJ is currently only available in English for premium subscribers in the US and Canada. As of February, the DJ feature is still in beta testing.

“We have some really cool new features coming out across the board,” Jernigan says. “We’ve got some really dope stuff coming out.”