Domo Arigato
A heartwarming story of an icon overcoming challenges...or a Trojan Horse for a trillion dollar industry?
Before I dive into this week’s topic, I want to make clear that I love Randy Travis’s music. In the late 80s I used to spend most days hanging around with Evan Dando in Boston listening to music, and we often chose country radio. Randy’s late 80s hits were “turn it up” moments for us, and we loved to learn and sing those great songs straight from the radio. I have fond memories aimlessly driving around in Evan’s mom’s station wagon on the North Shore singing along with all those great tunes, the foundation of a lifetime of appreciation of his work. I had other mainstream favorites in those formative years - Alan Jackson, Clint Black, George Strait - but it was that string of mid-to-late 80s Travis hits that pulled us in.1
I briefly met him once a few years ago at my friend Rory Feek’s home concert venue. Randy was the special guest that night, sitting on stage while Rory and a few friends sang his songs in tribute. He is obviously quite physically compromised from his 2013 stroke, including his ability to speak let alone sing, but he is also very present and aware. He beamed with gratitude and basked in the audience’s love that special evening. Clearly he would have loved more than anything to use his rare and unbelievably comforting and resonant voice to connect with the audience. It was both heartbreaking and deeply affirming.
I’ve followed the viral story of Travis’s new single, Where That Came From, with great interest. For one thing, it’s a convergence of all the stuff I love to think and write about: country music, the idea of authenticity, tech innovation and disruption, and music industry messaging and marketing, among other juicy topics. You’ve probably heard that his longtime producer Kyle Lehning used an AI voice model to facilitate a lead vocal track built on a guide vocal by singer James Dupre. If not, I’ll post the feel good piece on yesterday’s CBS Sunday for reference.
Before the CBS feature ran, I posted on social media to ask what people thought about the track and the technology. People have very strong opinions, both positive and (mostly) negative. Some strongly condemned the effort while others became quite defensive, asserting it’s offensive to suggest there’s anything weird or wrong. The overwhelming feeling, however, is discomfort with the whole thing. The quality is fantastic, and it’s a beautiful thing to hear that awesome voice singing a great song, produced to sound like his classic records. But still…
I’ll commend Warner Nashville that Randy Travis is absoultely the perfect use case to introduce this fast-emerging tech to an audience. He’s an icon with a huge, traditional fanbase that isn’t likely to be particularly tech-forward. He’s physically compromised but still living, cogent, and by all appearances able to make cogent decisions on his own. And most importantly, this is a rare use of generative AI that does not infringe anyone’s rights in any obvious way. There’s a market for new work, there’s a touching human interest/redemption story, and it isn’t infringing or illegal. It’s a guaranteed viral story (if not track) that places Warner Nashville at the cutting edge of a transformative technology that is about to take over the music business in ways we can’t yet really imagine.
Chris Lacy, the Warner executive interviewed in the CBS Sunday piece, lays out the mission as follows: “What would AI for good look like for us?” She doesn’t clarify who she means by “for us,” but I’ll assume she’s talking about her own company, Warner Nashville - but she could just as easily mean the music business. The mission itself - doing good - is an implicit acknowledgement that the tech can also be used for evil. As Warner Nashville pursues this heartwarming use of generative AI, you can be assured that the door they’ve opened leads to a very slippery slope that is being greased up and iced down by all manner of unsavory characters both inside and outside of music.
Sure, gen AI can be used for good, neutral, or evil ends. I’ve addressed my strong views on models that are trained on IP protected works without transparency, authorization, or monetization. But this is fully authorized by all rightsholders, apparently totally non-infringing! This is a rare example of a gen AI play in music that is legit. Does that mean that all legitimate uses of this tech are AI for good? What if the estate of a long-deceased iconic singer, relatives who never knew the artist but stand to make massive profits, decided to license their voice to the highest bidder? It’s easy to imagine how that would be horrifying yet fascinating that would be. Guaranteed virality, and totally legal and legit. This will happen, and it will be soon. Now that this sort of quality is accessible, it’s likely to hit all at once. Much of it will be infringing, but some of it will be legally defensible. I don’t suspect Warner of intending to soften public opinion for the onslaught, but I’m afraid it has a Trojan Horse effect.
Lacy’s next quote is even more nuanced and difficult to unpack than her “AI for good” comment. She says, “It’s Randy Travis - There’s no reason he shouldn’t be able to make music. To deprive him of that? It’s unconscionable.” But here’s the problem with that quote - nobody is trying to deprive him of the opportunity to use a voice model to imitate his vocal style. Using the tech is (along with his label that owns the recordings) his right, his choice, his and his label’s commercial opportunity. What some people are trying to deprive other people - and companies - of is the opportunity to do things that are either arguably illegal or just gross. It’s tricky to hold up this lovely, feel good use case as an example of all the good we can do with this technology because what are the other good use cases? More Randy Travis songs? Is it unconscionable to deprive anyone with the right to approve any use, or is it only unconscionable because he has a disability? What if the famous singer’s disability is no longer being alive?
This space is truly dangerous. With voice modeling, music is the tip of a very large iceberg. There are countless examples of people using this tech for horrifying crimes that destroy lives and livelihoods. Before we showcase to the world the incredible quality achievable by these models in the hands of professionals with rights clearances, maybe we should sort out the legal framework for managing the onslaught of deepfakes, frauds, and just plain old mediocrity that we’re about to have to endure on every digital platform we use? We still don’t have federal law on NIL/Right of Publicity. Cudos to the Tennessee legislature for getting ahead of this on behalf of celebrities, but what about the rest of us who are vulnerable due to something as mundane as being on a podcast or having an album on a streaming platform?
For the handful of folks on my social media who argue there is absolutely nothing wrong whatsoever with this use and it’s wrong to suggest otherwise, I respond that it’s risky to release high-quality AI-generated vocals before we have a sound legal framework, or even a good plan, to deal with the fallout. There is absolutely nothing wrong with the completely isolated use of a gen AI model to emulate Randy’s voice in a setting where Randy, his family, his longtime producer, and his label are all involved and approving. Nevertheless, this is a rare instance where it is not only legit from a legal standpoint, but also from a “use AI for good” standpoint. That’s why it’s a stroke of sheer marketing genius. But I’m afraid there isn’t a very deep well of other use cases like this one.
Challenge for Lemonheads fans - I think Evan has covered at least one of these before. Help me locate please. Forever and Ever, Amen?
I remember a lot of chatter a few years ago about the idea of virtual/CGI Marilyn Monroe's or James Dean's appearing in new movies. All these years later, I couldn't have imagined the music industry would get to the finish line on that first. I believe the idea even back then was to use an impressionist of some sort to do the voices on these new movies of the deceased. This vocal AI technology seems to have come a very long way in the blink of an eye.
Personally, I cringe the most at the idea of someone giving a "new" Elvis album this treatment. I hope they respect the more impactful and substantive work he did and not just go straight for the jumpsuit campiness. That said, it might be kind of cool if someone can concoct a "new" Sun Records-era Elvis album with some integrity to it. (Obviously, the estate would have a lot to say about that, but it doesn't seem so unlikely now).
John, this is a helpful contribution to what is sure to be a prolonged and thorny cultural conversation. I sure appreciate you adding your voice to it though. Keep it up.