Luma AI's Caroline Ingeborn on AI Video Generation, Creativity & Storytelling | AI Rabbit Hole 2025

You can take these two middle ones. Yeah. Hey everybody, I'm Josh Constine and today we're going to be talking about the future of creativity. We have one of the most incredible founders of a company, Luma AI, that is revolutionizing the way that people generate video both as a consumer app and as an API. And we have Caroline here, the COO, and she was actually previously one of the leaders of Toca Boca, one of the most popular apps in the world for children. I think there's like doing like hundreds of millions in revenue. And you know, it does feel like there's this throughline from finding how you implement and sort of activate childhood wonder and that sense of curiosity into what you're building now. So, I figure we just start with an easy question. philosophically, how does humanity adapt to the fact that we cannot believe what we see anymore?

Just real simple, just to start off. Philosophically, how does human It's a big question. I mean, I think that Okay, so I have a I have two kids and my nine-year-old I was looking in a magazine the other day. I love magazines still. So, I was looking and like she's next to me and she's like, "Mom, that's not AI." And I was like, "No, it's not." And then we she's like, "I think that is AI." Like whether or not that is AI or not AI, her mind is thinking like there's a generation. I don't teach them about AI or not not AI. Either YouTube or their teachers are doing this, but it's like that concept is already taking shape in the next generation. I think that we like I would like visual AI to get so good so that it feels real. I think it because that opens up for storytelling that it's really interesting.

It opens up for play and games and interactions that are really interesting. Yeah, it feels like, you know, when I went to school a long time ago, it was like, oh, you have to learn how to use Google and Wikipedia. And now it feels like we need AI literacy classes just to help people understand not only how to use these tools, but also how to protect themselves from misuse of these kind of tools. And that kind of gets to my deeper question. You know, we have a tendency to understand like if you write something, maybe it's not true on the other side. And yes, of course, there were special effects and, you know, big Hollywood studios that could make something make a dragon in Game of Thrones look like it's real. But generally, if you saw something in video, you assumed it was a real thing.

And especially as we get closer and closer to perfect photo realism, that ability to detect that difference of, you know, dissolves. So, what does that do to human beings? Like, how do we adapt and change our own behavior? Do we need to, you know, sort of verify everything and just assume a level of skepticism that we don't right now? And what are the kind of maybe positive impacts of that as well as the negative ones? I think it would be really sad if we need to become more and more and more skeptical. Maybe we do. I think the the positive side of what you're talking about is like around like only Hollywood studios having access to like the level of making dragons and now like this I don't know if democratization is the right thing, but it's it's really about permission. There are so many stories like they're literally drawers of stories that are not being told that are great stories that would be fantastic if they could be told with video.

they're just not being told because like they cost millions and millions and millions of dollars to make. but if they cost less or if more people had access to tell those stories then us as consumers and I love stories we could actually enjoy more stories and there are less I think gatekeepers sitting and deciding like the story and the story and yeah then there's going to be stories that are maybe not in this drawer of great stories that are being told that are in a drawer that of not great stories that are being told but then it's less is up to us consumers to decide what is what we prefer. to watch. Yeah. I mean, it feels like video is the most natural way for us to storyteller. It's the closest to what our own perception and interaction with the real universe is like. And historically, to be able to make great video, you either made something super scrappy on your phone, or you needed to raise a ton of money, hire a huge team, manage sets, and you know, storyboards and do all these different pieces of the process, which meant that it was kind of compounding the skills that you needed.

couldn't just be good at videography or coming up with a beautiful idea. Sort of similar to a hundred years ago, you could just be good at playing guitar and that's what it meant to be a musician. Then you had to be a good actor for MTV and look good on the Ed Sullivan show. Now you have to be an incredible data scientist and merch designer and everyone's friend. It feels like now, but in video it's almost going the opposite direction where you used to have to do everything and now finally you can do fewer and fewer things which hopefully unlocks new types of storytellers. Which brings me to the question of sort of what is the how does the nature of creativity change when this happens who you know what skills are are more relevant and what how does that change the way that we interact with the world and the media we consume so I think it's like if I go back to like what now feels like a really long time ago when I joined Luma which is less than a year ago we so a thousand years ago in AI time thousand years ago in AI time we launched dream machine and which is our now we call the video the video model Ray, but Dream Machine was like our cluster for product and video and everything that we did and and that was the first like widely accessible video model out there and it was like millions of people found it immediately which was fantastic but then it was like what could you then do and it was like explo like you could do you could make video and that was the big thing but after like I don't know 500

to 10,000 videos of explosions. It's like, okay, this was cool, but like now what? And then during last fall, we started seeing like more and more storytellers, advertisers, filmmakers, they were trying to tell stories. But the way that they would do this would be like, okay, I now have a story about a a girl that is walking from her mom to her grandmother through the forest and somewhere in the middle she she meets this wolf that eats people. And so if I'm now a creator and I want to use Dream Machine or any other tool to tell this story, I will start generating thousands of videos. And back in the fall, they just were not good enough. So then as a storyteller, you needed to start making all these shortcuts. So like the mother was not a mother. She was like, I know a raccoon and the wolf was a snake and the girl was a bear and like somewhere in the middle when you start like in this like slot machine of things happening, it's like there was some kind of story here, but now there's like animals in a forest and they're running around.

So there was like really good efforts of telling a story, but the stories were like kind of watchable. Like they were more interesting than the explosions in the first era, but they were still not stories. Now what we're seeing is like we're seeing stories. And I think that the the big reason is like the models are getting better, but there's more people using video AI and that means that more talent is using it. the models are getting better and image to video is getting really good and that is fantastic because storyboarding is like a traditional way of telling story and this old saying of I don't know what it's exactly and it's like one image carries more than a thousand words like that is true in the interaction between image and video in the AI land as well so stories are getting better and I think that that is like a really big part of the future of visual AI Yeah.

I mean, it sounds like previously in a lot of artistic creation and craft in general, there was this concept of measure twice, cut once. You know, you want to make sure you know exactly what you're doing before you do anything. Feels like we flipped that on its head now. Now it's generate a thousand times, publish once where and you know, instead of trying to say, oh, we need to know perfectly what we're doing before we even start, we can just bias towards action and learn from what comes out of it. Maybe the story changes in reaction to what you find the generated video gives back to you. And I think this relates back to a larger concept of what is AI due to us and how does it change creativity. You know, previously we were always the creators. And I think now we are the curators and we've moved from being the question answerers to being the question askers.

And so it's not about what can you make and how hard can you execute in all your hustle as much as it is do you know the right question to ask and you know can you do enough iterations that you get to better to asking even better questions over time. Yeah. And I think it it is like do you have a good story to tell because I I do think that that is like very human and and then it's about the ability of like what tools how f quickly can you adapt and how can you translate your story through these new tools. Yeah. Finally we're like learning the language. I mean you remember the early days of the app store when it was like oh cool like lightsaber app or it looks like I'm drinking a beer and that was the best we could come up with.

And then a few years later, we got like Uber and Instagram and real apps that took advantage of this. And it feels like we're kind of learning the language. Meanwhile, the models are getting better. You know, one of our consumer video investments at SignalFire called Nim video, you know, they were doing well until suddenly the models got way better in December and growth just exploded through the roof. And now it's such an easy way for anybody to be able to make faithfully put in a prompt, get something that is accurate to the prompt and actually just looks good. But that kind of brings me to my another question which is you're both building a sort of consumer product with dream machine as well as an API that seems like very ambitious maybe a split focus which a lot of investors don't love like how did you come to the choice to do that and what advice would you give to other founders in the audience trying to think about should I make just purely you know an enterprise product or API version or can I build a consumer version which is almost acts like a front door which helps me grow my top of funnel for potentially monetizing through the enterprise business.

So I I think today where we are at as Luma as a company, we're very much I wouldn't say like a broad consumer product. That is mostly because the technology isn't there. I think when the models as the models get better, there's going to be an explosion of consumer products built on top of video models. so we focused on creative professionals and for what that means is like some of them they want to access through a product like dream machine some of them like VFX artists like they're very technical they want an API this also means that we are selling our we could choose not to do that but we have decided as a company to say They're like, "No, it's better that we do that and we see that and we make our products standalone so good so that they survive regardless." And I think that has served us well so far.

but it does require a lot of intentionality and a lot of focus on where you do what. It's kind of curious because when I think of professionals, you think of maybe they want even better quality. like they're not comfortable with a little rough edge if they're actually going to be publishing this, putting it on TV, making ads out of it. On the other end, maybe you think that they are more willing to endure friction because they get paid for this. This isn't just a fun hobby like a consumer. Consumers I historically think of as like if it's frictionful, we just won't use it versus an enterprise will kind of throw bodies and budget at a problem until they solve it. And so, yeah, I'm I'm always fascinated. It's like does the models are the models not being good enough already? Is that better for consumer versus better for enterprise?

Because I think a lot of companies out there are wondering, hey, if the models have still feel like they're just on the cusp of being good enough to do this, what market should I be focused on mid-market, top of market, should I be going down for like SMBs and creators, prosumers? I I mean, I think anyone being it is say like on the prosumer side to a large enterprise side, anyone who is paying us today, they're getting value. I think that's sort of like and and when they don't they don't but but but it's a how the platforms that the content is sort of that the stories are told varies a lot we are not we see some of like I think there is some some large studios some large advertisers that have publicly shared work that they've done and that is really exciting but that has required them to be super curious to like spend a lot of time to understand the new technology and what what I'm seeing that is interesting is this new careers being born on the proumer side that are then connecting with the enterprises.

So now we have AI creators that are and maybe some of you are here in the audience but but that are like they they came in and they started using the proumer side for two different reason like two different sort of personas. One was people that ca like they were working as designers, as creative directors, as VFX artists, and they were like, "This is new technology. I've always been an early adopter. I'm going to I'm going to try this out." Or they were people that always wanted to be in this first group, but for various reasons, they never ended up there. And they said, "Whoa, here's a new technology. Maybe I can become that thing that I wanted to be." And the awesome thing is that like I see people in both of these groups being very successful. I also have like multiple sunshine stories popping up every week of people in the second group that are now getting to like have the career that they always wanted through these tools.

And so that's really like I see them either joining big enterprises or creating their completely new IPs for what you say, "Oh, my sunk cost doesn't matter. I just want to get great at this tool." Suddenly you can get so good so quickly that you can go from it being a hobby to it turning into your career in a way that previously you were going to have to dig your way out of you know competing with people who've been videographers for for 50 years. And so finally you know do you have a request for startup like you know other than just making traditional create creative stuff or competing tools what do you hope people do with these APIs now that we're finally at the point where the infrastructure is good enough to make new amazing things. What do you want to see people build with with new generative videos of?

So like I mean this is maybe like a non-answer I'm going to give because I could say like it's so interesting to look at the gaming side and all and and there are a lot of like very interesting if I look at anywhere where video exists today all of that is going to be more and going to be better. But if I look back at all of the historical like technological shifts, like what excites me the most is the thing that we do not know. like we had no idea that it would be interesting to sit and watch other like everyday people like blow their hairs on YouTube. Like that was not a thing that anyone would have said like I think that's what's going to happen when YouTube comes. like we had no idea and and that ended up like being one of the early things they saw at YouTube that people were interested in a doing which was really weird and b watching which was equally weird.

and I that is exciting with this like new space like what's going to be built like I actually don't really know. but that excites me. I haven't seen that yet though. I would say I've seen a lot of like these people are gonna need this and these people are going to need this and these people are going to need this. But something weird and exciting is going to happen. I love that. So just to quickly recap, you know, we talked about that these new tools give everyone the permission to create and there are so many untold stories out there that can finally be unlocked and maybe instead of thinking of being a perfectionist and measuring twice and cutting once, you generate a thousand times and publish once. And you know this tech is is you know isn't just for consumers and instead said it's unlocking careers for people who maybe previously had this like latent creativity and even if you think that your product isn't good enough that the AI isn't there yet you can build for the enterprise because as long as they're getting real value out of it they will endure the friction and they will wait for the models to get better for you.

And to me, I think the most exciting thing is when you suddenly allow such a huge number of people to interact with a type of creativity that is so natural to us, our own visual perception of the world, that the truly weird and wild things are just on the horizon. So, thank you so much for sharing with us today and thank you everyone here rabbit hole. Thank you. Thank you.