NVIDIA’s Vision for AI, Music, and Hollywood

0:03Dan Runcie: I'm Dan Runcie. Welcome to Trapital. Today, you're going to hear a conversation that we had on stage at our Trapital Summit about how AI is reshaping the future of music, media, entertainment, and more. We're joined by Richard Karras, who is the VP and General Manager of Media and Entertainment at NVIDIA. You may be familiar with NVIDIA's work powering the AI revolution that's been underway for several years now, but NVIDIA also does a tremendous amount of work with leading companies in the space, including Universal Music Group, Runway, Getty, Lovable, and more. Their job is to help make the process more efficient and help these companies think about what's next, How can they integrate and leverage all that's possible with AI, but still respecting the underlying IP that is core to their business?

0:54Dan Runcie: So Richard and I talk about that, how the partnerships look at these different companies, his take on why sometimes the messaging between what some of these companies may say publicly may look different than what happens behind closed doors, what he predicts for the future, working with the company CEO, Jensen Huang, and more. So I hope you enjoyed this one as much as we did. Here's my talk with Richard. This episode of Trapital is presented by Symphonic Distribution. If you've been following the deals in music rights and catalog sales, you know that there is plenty of volume and interest in this space and it continues to grow.

1:33Dan Runcie: And that's one reason why Symphonic has launched Next. Next works with independent artists, managers, labels, and rights holders who want to explore the value of their music catalogs. They do that through offering flexible advances, partial or full catalog sales, strategic partnerships, and long-term distribution support. And all of that is backed by the infrastructure that Symphonic has to manage, grow, and optimize catalogs for a lifetime. Whether you want to fund future releases, expand your business, or plan the next phase of your career, Symphonic will offer transparent and collaborative solutions to unlock and manage your catalog value with Next.

2:14Dan Runcie: And because they're one of the leading music distribution platforms, Symphonic will offer a unique perspective on catalog performance and help you make the most informed decisions about your next step. If you want to learn more and explore your options, go to symphonic.com slash next or visit the link in our show notes. I think to kick things off, during one of the networking breaks, I was talking to someone in the audience who did candidly ask me, so what exactly does nvidia do in music because they understood the compute and the chips but what does your role look like in media entertainment i felt like that would be a good opportunity to kick things off because i'm sure other people may have had a similar question well first of all we love music and so we're fans of it all but nvidia is a company that solves problems and historically in the media and entertainment space a lot of those problems have been the visual side of things, to render images that look photorealistic and how long that would take.

3:10And now that's a real-time capability. With music, there's been challenges of accelerating the tools that are allowing you to mix music and things like that. But today with AI, there's so much more that we can do. My teams, we're focused on media and entertainment. So primarily it's been audio and video and such, but now bringing in audio into the mix, there's a whole lot of opportunities that we're seeing there.

3:33Dan Runcie: You've talked a lot about digital media as well. And you've said that AI will be one of the most transformative things in this space. And I know a lot of people likely agree with that and understand, but if you could talk a bit more tangibly on what that looks like in terms of what you think will be possible. Sure. I think the first thing that's important is there's a lot of people out there that think AI is a thing. And like a noun, you add AI onto your existing thing and something happens, right? And I look at that as like buying a synthesizer and hitting the demo button and think that's all it does.

4:09AI is so much more. It's an action. It's something that you can utilize. It can be an assistant for you in the workflows that you're doing. You can use agents to help you explore, experiment, things like that. It can be a tool to help you navigate content in your archives, things like that. You have to understand that AI has such a depth and breadth to the capabilities of it. It's not just about somebody hitting that demo button and saying that was made with AI. That's kind of boring. But when somebody says, I was able to do this at a faster case than I would have in the past, at a higher fidelity that I would have been able to do in the past, and I did it on my own because I had AI as an assistant on that process, that's transformative.

4:53Dan Runcie: So for you all in your work, I then assume that speed and autonomy are like two important metrics that you look at and say, okay, if we achieve those, then that is a success. Absolutely. And of course, artists' rights are paramount to all of that. Like the work that we're doing with UMG, we're exploring how they can go through their millions of songs in their archives and get information from those songs that can then be utilized for new purposes, new business ideas while retaining the artists' rights. So I'll give you an example. They have about 10 million songs that we're working with them on with our research team.

5:2810 million songs for a human to go through and get information about would be a lot of humans and a lot of time. And that information would probably be just based on the title and the metadata, perhaps, about the song. You wouldn't get things like the emotion of the song or the timbre or the instrumentation and all those other things. Our research model, Flamingo, is going to allow the researchers to understand all of that music in their archives and be able to bring back tools that will allow somebody to say, I want something that's a happy song, that's this long, that's all those kind of parameters.

6:05And then when you think about that in a business model, you think about somebody that's making a film and they want a piece of music that's 30 seconds long to fit a particular scene. And I wish I could find something from Aretha Franklin that would fit that piece, right? They could simply prompt that in, get that thing, use that piece. The artists or the estates in that case gets the credit that's deserved and they get compensated and things like that. This is a new business model that we're just exploring. We're at the frontier of what do these things mean for both the artists and the use of this content?

6:38Because I'll bet there's not a lot of people that have gone through 9 million or 10 million songs trying to understand what they all do. But now you're going to have that at your disposal. And that's just one element of how AI can help in those kinds of scenarios.

6:53Dan Runcie: So for you all, given the partnership that you have with Universal to do this, and it seems like their insight, their information is helpful with your music Flamingo model. Are you looking to have similar types of partnerships with the other major record labels and some of the other larger independent music companies? Well, certainly with the companies that the labels work with, typically to understand how we work, we don't have an end user product. So to work with a company, somebody usually brings us a problem, a challenge that we look at and we work with them and say, if we solve this for you, we could solve this for many others in the industry as well.

7:29Right. There's a lot of other companies that have archives and don't want to do things like that. So we approach it in that context. What's good for you has got to be good for all, but we're not a commercial company that somebody comes to and says, I want to partner with you to make a new album for this artist or something like that. As much fun as that would be, that's just not our business. We would instead rather work with you and the companies that you use to make those albums to make sure that they've got the latest and greatest technology in those software products that are being used.

7:59Dan Runcie: And I think a lot of this too is about the messaging as well around this and like what people may say publicly, which, and I know as you've talked about before, can sometimes be a little bit different from what they may do privately and even how sometimes those companies may act. And it'll be good for you to speak on that a bit. Why you think there is that disconnect and where you think that may go. Especially like in the film industry, I can talk about it because having been in the film industry for so long. When people don't understand the whole aspect of something and they immediately jump to an opinion out there, it's very reticent for studios to then come out and try to fight with them and things like that.

8:36It's better for them to take a little bit of time and let the learning happen. So for example, in film, especially a high-end production film, AI has been around for a long time. This is nothing new. We've de-aged actors for years. We've used AI to do rotoscoping, all kinds of things like that. And no one's ever questioned it, right? It wasn't until somebody saw images that were created with AI or using AI, and they said, wait a minute, where'd that come from? And the researcher said, we just trained it on things because we wanted to see if it was possible. So they first saw what a research project was, and then I tried to apply that to a commercial sense.

9:13And that wasn't the intent at all. It was more of the researchers saying, can this be done? and if it can be done, what could be done with it. Now studios are using it in a lot more use cases. And of course, again, the key point though is they're not using it to replace an actor or an actress or things like that. They want to use it to enhance.

9:37Dan Runcie: Imagine you're running a major record label. Let's say you're Elliott Grange. You run Atlantic Music Group. You have five successful artists on your roster. Bruno Mars, Don Tolliver, Alex Warren, Pink Pantherus, and Stella Lefty. All at different stages of their careers, but with a limited marketing budget, where do you invest your dollars? I put this same exact question to Flo, and I asked it to analyze each of these artists' streams, radio, playlist, audience demographics, and social performance. And let's go one by one here. It's really fascinating. Take Bruno Mars. His monthly listeners on Spotify are actually declining, but his followers are increasing.

10:20Dan Runcie: A lot of this is what they call post-release normalization. He released an album earlier in 2026 called The Romantic, so it's more a sign of that, not necessarily weakness. But Flo suggested to focus investment for Bruno on Latin America and Southeast Asia. That's where his audience is already concentrated. Let's look at Don Tolliver. He has the strongest streaming acceleration of the group, but 91 % of his radio airplay is currently in the US. There could be an opportunity to expand that internationally. Let's look at Pink Pantheris. She has exceptional engagement economics, but the challenge isn't connecting with fans, it's reaching more of them.

11:01Dan Runcie: She could be a strong candidate for incremental marketing span. And Stella Lefty, she's breaking out. But 70 % of her playlist reach comes from editorial. Nearly 95 % of her radio play comes from her one hit song. That's momentum, but there's also a huge concentration risk. The priority is converting some of that attention into lasting fandom. And that breakdown right there for those five artists, that really impressed me. And that's what Sharp Metric's flow can do. It doesn't just tell you who's winning. It doesn't just tell you their streams or their followers. It helps you identify where the next dollar can make the biggest difference, and that's the power of turning music data into strategic intelligence.

11:43Dan Runcie: If you want to check out Flow for your own roster and see what insights you can pull yourself, then click the link in our show notes. All Trapital listeners can use our promo code before you sign up for Flow. And please let us know what it comes up with. I also have to imagine that for you specifically with music and with film specifically, that makes sense. I know you worked at Lucasfilm and elsewhere in the space, But in music, though, do you feel like you also see a bit of that disconnect where the public messaging from some of these companies is different from the conversations that you may be having directly with them?

12:15Yeah, I think that, again, it comes from my perspective of fear and uncertainty of saying the wrong thing and getting people upset with you. Because I think AI, when it first came out in the generative AI sense, got a little bit of an unfair rep because of the researchers were the ones that brought it out first, not the commercial sense. I think if a commercial sense brought it out and said, we can help more people make more music to bring more content to the people that want it at a lower cost and help democratize it, people would have reacted a lot differently. But when somebody sees something and says, wait a minute, that was made by a computer or something like that, they have a visceral reaction to that.

12:55So I think we saw similar things in the music industry. Remember sampling? Sampling came out and everybody was very up in arms and major artists were actually quite against the sampling thing. Now it's just a regular part of the toolkit and sampling gets used all the time. Synthesizers, same thing. Drum machines, I'm a drummer. I remember when drum machines came out and everybody had a real bad reaction to drum machines and now they're just another tool. Use it if you want, don't use it if you don't want. And I think that those people or those companies that might say something privately and then publicly have a different kind of position, I think they're just trying to play it safe, right?

13:33They just like, let everybody get this settled in and then it'll just be another part of the fabric. I think that's really all it is. That's going to broaden that footprint of capability for artists. That's a good thing, right?

13:45Dan Runcie: And I generally do think that that is a good thing. It's funny, though, because sometimes I do hear the flip side of that, where some people say it's already so cheap to make music. It's already so easy to make music. Do we flood it more by making it even easier, especially with these tools? And I know the people that work in storage are complaining about, oh, we have all these AI tracks and these music streaming services. I'm not saying that necessarily those statements don't have their nuance, but I'm sure you probably heard some of that. Oh, of course. But again, look at the dynamics. Over six and a half billion people a day interact with media of some kind.

14:20And a lot of that is audio, right? Of course, there's going to be a lot of noise in there and stuff like that. There's always been, right? But there's also going to be a lot of gems that come out of that. And I think one of the things that these tools help you do is to find those gems and the things that you like. I use an agent to help me explore new music because I'm so locked into the stuff that I grew up with that I love so much. but I go, I don't listen enough to new music. And a lot of times I just put something on that hasn't moved me. But if I can search and go, if I like these kinds of things, are there artists making this kind of music in a similar vein that I should know about?

14:57Oh yeah, there are. Here's some artists and stuff. I think, yes, there will be a flood of a lot of stuff. And yes, there will be better tools to help you navigate through all of that to find what you're after.

15:09Dan Runcie: And I know we've talked primarily, at least so far, about music, TV, film, but with your gamut of what you look after, there's also partnerships that you have with companies like Adobe, companies like Getty and Runway and others. A bit different with those, but I think they still maybe have some similarities. So can you talk about some of the infrastructure goals there in terms of what you all offer and how it may look a little different from some of the IP? We've worked with Adobe for decades, Runway, now eight years old, I think. We've worked with them. And again, how can we help them solve it better?

15:43Historically, like with Adobe, can you help us make things run faster? Can you help us have a higher fidelity of color and things like that? With Runway and their incredible models that they're making, how can you help us make these models run faster and be more accessible to people? But there's a lot of other companies too. We're exploring ideas with Roland. Every musician in the world, I think, knows Roland. And wait till you see some of the things that they're talking about doing. Because they're like, how can we take some of this AI technology and our understanding of augmenting sound through foot pedals and devices and things?

16:19And what happens if we merge these together?

16:21Dan Runcie: And I know that with a part of your role in general, in order to do this, you need to be at the frontier of technology, always paying attention to where things are heading. But I'm sure that you probably also decipher between, OK, where's the place that we need to be early as a company to make sure that we are serving the customers and ultimately doing what's best for NVIDIA, but also where you may want to pump the brakes to either see, OK, how is this trend going to play out? How strong is this going to be three, five years ago? How do you distinguish between that? That's a great question. That's one of our challenges that Jensen gives us all the time, because everything looks wonderful at the buffet table of opportunity when you're there and you're exploring ideas.

17:01But he will challenge us every step of the way. If we solve this problem, how big is this opportunity? Are we going after a zero billion dollar opportunity? And by that he means, is there nothing there today that could be a billion dollars or more if we solve this problem? And that helps us to make decisions. And plus, listening to our ecosystem partners is the other part of that algorithm that, again, it really comes down to like all these things would be fun to do. But if we do these three, they're going to have a rippling effect that goes much further. And that's just something that's always ongoing because we see lots of times, lots of cool things that I'd love to go spend time on.

17:41Maybe I'll do that on my own. But in the company's aspect, it's like it's got to have that big footprint. And that's what's driven the company throughout everything that we've done. And do you look at that retroactively as well?

17:53Dan Runcie: Let's say, for instance, the metaverse. I know that was a huge focus for you all at one time and the landscape has clearly changed. I think with that same framework, are there any takeaways you have from that in that same type of way of, okay, what made sense for NVIDIA to do at the time or, okay, what maybe we could have read that differently? I think the metaverse and the marketing is what got a little bit out of hand, like the same way multimedia got out of hand years ago and things like that. But is it happening? Absolutely it's happening. Is there a multimedia? There's all obviously been multimedia.

18:23The term has gone away, right? The metaverse term got overused and oversaturated without understanding. But what was it really about was creating digital twins of a physical environment in a digital environment. And it's happening everywhere. It's happening in automotive manufacturing, where digital twins are built down to the nuts and bolts of a factory and taking place synchronized to the digital world so that decisions can be made cost-effectively, safely in the digital world before implementing it into the physical world. Same holds true with autonomous driving, with robotics, all of those things, all of which stemmed from all that.

19:00We just don't call it the metaverse anymore because everyone, that name got so overused as, oh, it's multimedia. It's like people would laugh at you if you said we're going to do a multimedia presentation, but we had graphics, we have sound, we have multiple medias, right? So I just think that sometimes the marketing gets ahead of itself and things like that. And that's what happened there.

19:19Dan Runcie: So you feel like the work that you're doing is pretty much similar to then, even when the titles and the roles reflected a bit more metaverse, omniverse-centric language. Yeah. I mean, omniverse is the world's most advanced toolkit that's been made at NVIDIA that is a toolkit for developers to realize those things that I've talked about, whether it's robotics, digital twins, all kinds of things, the physical AI capabilities of these things. All of those are components. So we don't ever chase after something just because it has a cool name. We actually do all the research and listening to our partners to see what problem are we solving?

19:56How big will that solution impact the industries that we're serving? And if we invest in that, are the partners that we're working with going to invest as well? And then there's a better outcome from it all. The marketing is second or third, whatever. So a coined term of something doesn't motivate us as much as the opportunity should.

20:15Dan Runcie: Are there any type of companies within media or entertainment that you look at as an opportunity where NVIDIA can be helpful, but you haven't either been able to work through yet, whether it's because of the opportunity or just because the technology isn't quite there yet, but you hope that it'll get there? Lots of times, yeah. And I see that through lots of different devices. I wish I had more time to spend with the instrument manufacturers and things like that, because I think that there's a lot of areas where this technology could help in your music, in the device that you're playing, to keep it auto-tuned or to keep it harmonizing with other tunes.

20:50There's a lot of stuff that we look at, but again, there's only so much that we can do at any given time.

20:56Dan Runcie: So given that your role is both the partnerships you have with these different companies, but in many ways, they also may be looking to you for some informal guidance on, hey, what should we be paying attention to? What should we not be paying attention to? What's the trend right now that may often come up in those conversations that you urge to them or suggest that, yes, you should be following this right now? A lot of it is understanding the depth and breadth of what's possible with AI. Here, a quick example. I know we're going to be running out of time, but like software is hard to learn, right?

21:26Here's a perfect example of why agentic AI is going to be so impactful. A lot of software is really hard to learn. So you become a master in a certain area. You might be a master at producing or mixing or creating 3D objects or what have you. And you then get to rely on others to do other things for you. And that's costly and time consuming. I mean, your agentic, your agents that are AI driven can understand those software products for you to realize your creative vision. So you can say, I want you to go mix this way or I want you to go create a graphic that will go along with this. And as if you are directing your team, you're directing these agents to give you more creative options, more creative output.

22:08And that's going to be something that I think people are just going to be like, I never realized how much power AI gives me as a creative. where I was always so afraid that it was doing something that I was supposed to be the creative. You are the creative. You just have a lot more power at your disposal. I appreciate spending our time with us.

22:24Dan Runcie: My pleasure. Richard Karras. And that is a wrap. Thank you again to Richard for joining us. Thank you to Sean T. Smith for capturing the audio and video at this year's Trapital Summit. Thank you to G, Eric, and Abigail for everything you do to help make this Trapital episode possible. And most importantly, thank you for listening. If there's one person you know that would really enjoy Trapital and get a lot out of it, then send them a link to the show. Word of mouth is still the best way to grow. And if you haven't already, tap that follow button, make sure that you are subscribed on Apple, Spotify, wherever you get your podcasts, make sure you get the next one.

23:00Dan Runcie: Leave a comment, leave a review. That helps make sure that the algorithm does its thing and that Trapital reaches the right people. Thanks again. Talk to you next time.