When you go and watch a movie and you pay to go to a theater, why are you watching that movie? Do you care which cameras were used in making the film? Do you care about the software that was used? If you go there for one thing and one thing only, which is you want to be entertained, you want to like the story. This doesn't change the equation. If I can have 100 more people make 100 more movies that you like watching, then you will watch them. You will find the time to watch them because that's what you find valuable. I think that the dynamics and the formula and the equation that's changing is how many more people can do that and how many people can tell their stories. The way we value them will still be the same, like nothing will change. You as a viewer are going to judge it with the same values that you had before, which is is it a good story or is it not?
Welcome to Inspired. I'm Alexa Vonto. What if the next Oscar winner made their film entirely with AI? Well, Cristóbal Valenzuela is the co-founder and CEO of Runway, the company that pioneered AI video generation and is building world models that will power the next era of human storytelling. The company is now valued at 5.3 billion. Their models are being used by Hollywood studios, gamers, architects, teachers, robotic researchers, and everyday creators around the globe. In this conversation, we're going to talk about why AI video is an entirely new medium that requires an entirely new mental model. What happens when the cost of creating video content truly goes to zero? How these models are already acting as genuine creative collaborators and why every major technological shift has felt catastrophic in the moment and transformative in retrospect. Let's dive right in.
If you're running a startup, we know things like managing equity and communicating with investors are always top of mind. Carta has built a connected suite of tools that over 40,000 companies use to do just that. Their platform of software and services helps you lay the groundwork so you can focus on building. To learn more or get started, visit carta.com.
And with that, let's welcome Chris. First of all, Chris, I'm so grateful to have you on today. I know our good pal Howie from Airtable had us meet. I'm really happy to have you here, especially as a fellow New Yorker. So first, just thank you and welcome. I want to start from the beginning. One of the things I find so interesting about you, especially with this incredible business that you built that's sort of the intersection of the future in so many ways, I want to take a quick minute to rewind. You grew up in Chile and you moved to New York to join NYU's ITP program back in the day. Growing up in Chile, how do you think that changed you? How do you think that impacted you? It would just be so great to hear a little bit about that.
Yeah, happy to. And thanks for having me. Great to see you again. Yeah, you know, it's true. I grew up in Chile. I spent most of my time in Chile. I came here to New York almost 10 years ago as you were saying to study at ITP. Well, I think Chile formed me in a lot of different ways. I studied there. My family's there. My upbringing was there. And in a way, I think I had to bring a lot of that into the way I started to build Runway and the way I maybe helped shape the culture within the company early on. Chile, if you've ever been there, it's a very different country and culture from the states and from the US. And so it was definitely an adjustment period at the beginning, specifically within like speed and how things are done. Chile is very stuck sometimes in the past in some particular ways. And so yeah, changing and adapting to it has been probably the biggest lesson as I've been here for now more than a decade or so.
That's wonderful. If you think back to 2018, can you just in plain English give us a sense of where did the idea for Runway come? What was the aha moment? And in so many ways you actually built Runway well before generative AI took off on the main stage. You were extremely early. Let's start a little bit with where was the idea? Where'd it come from?
Yeah. So the idea, kind of the origins of Runway, date back to our work as the three of us co-founders were doing at NYU. We were trying to understand and research this idea of using AI and deep learning for creative use cases. And more of like, oh, there's one particular idea I want to explore was more about the concept of using and training large models to help solve or understand how they can be useful within creative environments. And that was like the premise I would say, I think more than an idea. We had like some boundaries and some context on which we thought was worth experimenting and building. And over the last 8 years and since we started the company, we've been relentless in making sure that that vision and the thesis still becomes true, which is using AI as a means to augment human creativity and human intellect. And I think we've gotten far further from realizing that in some way, but we're still a long, there's still a long way to go to be honest.
In plain English for everybody out there, what is Runway doing and what does it help people do?
Yeah. So, in plain English, what Runway does is we're an AI research company that's building what we call world models. And world models are AI models that can understand and interact with the world perhaps in similar ways that we do. And the applications of that technology and those models are broad. There's a lot of applications as you can imagine in film and media and advertising, which is where we're kind of growing the most these days. You can create these very sophisticated models that you can then deploy and be kind of co-pilots of creative minds. And so we're very much used these days in Hollywood and in media and everyone is using these highly sophisticated models to do things that they just couldn't do before. But the idea is also that these world models and these video models can have such a wider spec set of applications beyond media alone. And so what we do at Runway is figure out what other applications we can deploy the models into. And we now have use cases in robotics. We have use cases in the gaming world, in education and in many other kind of disciplines. And at the core, Runway is both training the models, creating the technology, innovating it and then building the products and the experiences on which millions of people can now use them.
Chris, give us a few examples. You just rattled off some categories. Give us some practical examples and use cases.
Sure. So you can think about basically three main use cases and I'll walk you through each of those in a minute. But the first one is with this idea of using AI or video and world models to create entertainment or create media. So if you're ever watching a movie or if you're going to stream something in your house, you are watching the content or the scenes or the videos someone either recorded or produced with traditional VFX CGI techniques and tools. Today where AI is, we are at a place in time where a model can simulate or generate almost all of those scenes. And so if you are in the business of storytelling, in the business of filmmaking, in the business of making media, because this also helps and works really well for advertisers and basically for any other content creator out there, you can use these models to do that at a fraction of the time that it would have taken you to do it traditionally. And so the more concrete use case is, well, this is a new kind of camera. This is a tool and a set of AI models that can help you tell stories visually in ways that you couldn't do before for high-end professional cinematic worlds and for casual creators. And so that's a little bit of the user base of Runway. We work with the best and greatest Hollywood studios in LA and the world. And we also work with just consumers and creators who are using Runway to have fun. So that's kind of one concrete use case that's actually getting deployed a ton these days, is just as a means to create video and content. A second kind of possibility is what happens when you start running these models in real time. So again, you can create movie-like content or really compelling stories, really compelling scenes. What happens the moment you start running them in real time? This is real time, right? We're speaking to each other. You're responding to my questions. I'm asking your questions as well. You can do a version of this with AI models these days visually, which is kind of surprising and kind of magical to be honest. And so there's a lot of interactions and things you can do. You can have personalized tutors, you can teach people about new languages, you can just have a game-like experience where you can chat with characters you couldn't do before. So that's another kind of compelling use case of our technology these days. And the other one is what happens when you start thinking about not only creating content and videos we humans want to watch because we find those kind of two categories entertaining. We can also create videos for robots to watch. What I mean by that is if these videos are good enough for us, they might be as well good enough for robots and robots can watch them not for the sake of finding this entertaining. They can basically use this as a way to learn about the world, learn about how the physical world works. In other words, you're creating data sets or training data sets. And there's ways of also evaluating how effective these data sets are. And so we are kind of focusing on those three main areas these days. This idea of async or narrative storytelling. This idea of real-time or interactive experiences and then the robotics kind of world physical AI aspect as well.
There's a lot that you just said and I'm trying to think about the best way to ask this question. So Runway's first models like Gen 1 and Gen 2 were some of the first widely used systems for video-to-video and text-to-video generation. I want to ask this very specific question which is technically speaking what had to become true in order to have that step function change that Runway created, which in so many ways you've now made video creation wildly accessible to a layperson. What technically had to happen for that to work?
Yeah, that's a good question. A couple things had to happen. I mean there's definitely compute and resources had to be in place for training models at that scale. And again, early versions of Runway's video models were decently small compared to where things are these days. But you had to have the infrastructure, the technical infrastructure to train those models. The second part was there had to be interesting algorithmic ideas around how you can actually deploy and train the models in the first place. And the second one, data, there needs to be a good amount of really well captioned and kind of curated and well-sorted data. And a lot of those things have been kind of progressing over the last couple of years. But I think over the last maybe four years or five at the most, things have converged in terms of compute availability, good algorithms and good data. And that's I think some of the results is the models that you just mentioned and hopefully more. And this is not only for video, this is just generally for the industry. So language models and coding models and all of them have benefited from those three variables kind of becoming better over time.
In the last few years, what has surprised you about the progress of Runway? What has been surprising to you, Chris?
There's a lot of things that are surprising every day. That's the goal of building some of the stuff we're building, you know. So, a couple things I mean definitely come to mind that I think have been kind of consistent patterns have been one, underestimating the speed in which things are changing. I think it's a common misconception to look at a model and see all the imperfections or the flaws of models these days and tend to think that it will never work because X, Y or Z doesn't work. And I've heard that for the last 10 years consistently and we keep moving the goalpost. People thought images will never be able, models will never be able to generate images and here we are. We haven't solved images, but we're very decently close to solving the idea of generating pretty much any image that you want. People thought video generation was 10 years away and it was not. People thought then you couldn't generate hands with five fingers and all of the things that I think we keep coming back to as to what models are not going to be able to do. We've proven and the industry, I think this is a collective research effort of many other people, have proven that the models should just keep on getting better. And so speed and underestimating the progress is perhaps one of the key lessons and surprises I've learned over time.
On that thread of kind of how people are maybe not thinking about this properly, where are people still using the wrong mental model when they think about this shift in video and how does that lead to maybe some of the wrong conclusions?
Well, I think it's rooted in this idea of how we've and industry experts have understood how video and world models are used is maybe coming from a place of comfort and a place of understanding of how the world works. And so when people who work in filmmaking look at video models, they tend to think of them in this very neat specific way of how every other tool fits within their workflow and way of doing things. And the reality is that for someone to truly get the best and the most successful outputs of these models, you need to rethink things from the ground up. It's like first principles thinking all over again and need to understand that it's less about making things 10, 20% better, is maybe you need to just rethink your entire work in the first place. And I think we made the mistake in understanding new mediums, we made the mistake many different times. Even with like when cameras were first around, people made the mistake of thinking them as an extension of painting and of course it's not an extension of painting. It's a different medium. You need to understand the world differently to use a camera and we've gotten used to that but that wasn't the norm back then. And so again, for me it's a new medium and so you really need to think from a different mental model how to best use the models to do what you want to do.
That's really well said, which is that the shift from painting to photographs was a different medium and now AI video generation is a fundamentally different medium. It is not you with your camcorder or you with your iPhone recording video. That's a great analogy of the mental shift that we have to think about. Maybe this is a good time to ask this question which is, and I saw you once give a talk at a thesis event about the future, which is why I'm so excited to ask you more questions about the future. If the cost of creating content goes to effectively zero, right? It's becoming, and it's not just text content, it's truly long-form, medium-form video content can be generated really to near-zero almost. What does become scarce and where does the value actually move? How do you think about that shift in value, Cristóbal?
Well, there's an argument to be said that there's no shift in value. Maybe I'll rephrase the question, but like, and I'll ask you. It's like when you go and watch a movie and you pay to go to a theater, why are you watching that movie?
Because I want to see the content. I'm going to see The Devil Wears Prada 2 in theater.
Why? Why do you care which cameras were used in making the film or that particular feature? Do you care about the software that was used? If you go there for one thing and one thing only, which is you want to be entertained. You want to like the story. The value you're paying for that ticket is not rooted on the technicalities or the complexity or the budget the movie had. You just want to go because you feel it could be a good movie to watch and you like watching movies, right? So the value is the story, right? This doesn't change the equation. If I can have 100 more people make 100 more movies that you like watching, then you will watch them. You will find the time to watch them because that's what you find valuable. I think the dynamics and the formula and the equation that's changing is how many more people can do that and how many people can tell their stories. The way we value them will still be the same. Like nothing will change, but now way more people are going to be able and are able to have the opportunity to tell you a story you couldn't do before, but you as a viewer are going to judge it with the same values that you had before, which is is it a good story or is it not. That I think will not change and we've proven that through the history of art that still is the case.
Basically, I'm gonna jump to conclusion. We will have an explosion of content creation because you put this incredible power of almost an LA studio into the hands of a normal everyday person. But your point of the quality of what's created, the story, whether it's interesting or not, compelling or not, won't change. But when you maybe, and again maybe you have a point of view, when you 10,000x the volume of content, does that change its value?
No, I don't think so. I think we just figure out ways of navigating and sorting through content better. And again, I'll go to an extreme example but the first ever storyteller was someone perhaps even in a cave telling someone else's story at night with fingers and maybe drawings on the wall and then suddenly someone was like, oh we can actually take the fire and put it in four other caves. That same person will be like, well it's too many stories. We have like four people now telling stories, right? It's like we need to, there's a massive amount of how many you can be in four places at the same time, like stop this, right? That same story has been told dozens of times where when publishing and books became something that you could do at an industrial scale. That same direction, like well how many books are you going to read? We have 4,000 books being published every single year. Do we need more books? Like does the value of book publishing will decrease? And I think the world is in a much better place when we're publishing millions of books these days. And of course you don't read them all, but we have libraries and reviews and sorting and we have way more ideas coming out of the world because the tools that would enable you to tell those stories have become more accessible to people. And I think this is kind of the same point which is it's another explosion of storytelling and the people who can tell stories. I don't think there's an optimal amount of content out there. That feels like a little bit for me like saying, hey, you know what? We're gonna stop progress all together. The amount of movies that we have right now and content is the optimal. We can't do any more else. No, I think there's like way more, the demand for stories and sharing ideas of the world and perspectives of the world is never ending. And I think hopefully more people will help to do it with technology.
Chris, when you think about how you set out in 2018, you know, eight plus years ago now, to build the product and then to build the business model. First, tell us what is the business model of Runway and how did you prioritize building this truly groundbreaking product which is anybody can create incredibly high quality generative video. How did you think about balancing a business model and a way to pay the bills versus getting the product right?
Well, I think those two things should be intertwined. If you're building a good product, then people are going to pay for it, then you can pay the bills, right? Like those two things should definitely be connected somehow. The business model for us is very simple. If we help you make something that you think it's worth, you will pay us and you'll help us pay for the bills for us to give you that. And so we charge for a subscription or for credits. But the value of the product comes from people being able to tell stories with it. And so that's why we host, for example, this film festival in New York and in LA every year to highlight some of the best creations people are making with AI these days to showcase the amount of creative explosion happening these days when it comes to Runway and to AI. And I think we want to see more of that. The business around it should be like, well, if we're succeeding at helping people be creative, then there's probably a nice and sustainable price we can charge for that.
When you think of some of the more surprising ways that Runway is being used, what are some of the ways that kind of surprised you to see the product in the wild?
Oh, there are many. I think I recently crossed a threshold where I have three brothers and the three of them have been using Runway every day for the last couple of weeks and that was like a nice surprising moment because they're not, one of my brothers, my brother works in media and advertising so I felt more of a natural fit for him to use it. But my two sisters are like an architect and a teacher and so I was never expecting them to use Runway and now they use it every day. Which is great. Like my sister using it to teach classes and help her students engage with content. My sister's like visualizing the designs she's been doing for her clients. And you have like hundreds if not thousands, probably millions of use cases that are similar to that which are things you might not even consider are useful because they're different from your day-to-day world. But they're very useful for many people. I mean we have use cases of Runway for music videos, for learning, for medical teaching, there's people using it for architecture, for design, for filmmaking, for many other things that I think we never thought of when we started but that's kind of the point. It means that these general purpose systems are becoming more usable as the models have progressed and so probably there's still a long way to go.
As these tools that you're building continue to improve and obviously you know there's sort of an asymmetric curve in terms of just the power and the potential. It's going to continue to get better faster. But what performs best today in the Runway algorithms? Like what's working extremely well?
What do you mean in terms of...
What type of video creation? Like give us some of the examples of things that are just kind of stunningly beautiful and powerful.
Well, I mean it's hard. I mean the models are very flexible. They can do a lot of different things and different art styles. I think long-form content is becoming more feasible. Like we used to create like single images, then shorter videos and then longer-form content. Now I think artistically it's less of a problem. I think the biggest challenges you can think of like consistent forms and styles not being solved but being very close to solved. What's the hardest these days is motion and physics and consistency like long-form content. That's I would say where we need to put perhaps the most research and time to make sure it works as it should.
Can you imagine a future where someone sitting on Runway like wins a solo Oscar? Like is that...
Yeah, absolutely. That's like hopefully a couple years away from now, hopefully less.
We were an early small investor in Suno AI and they, you know, people started creating literally award-winning content and music. So I imagine the analogies here are very similar. Are we there? Like could that happen right now almost?
Yeah, I mean I think the question there is not if it can happen, it's more how will we position and well prepare our institutions for that. Which is not a question of like if it can, it's like why and when, you know. And look, people are making incredibly compelling videos and stories that are fully synthetic or fully generated or a mix of real live action and also generated and it's just really good. Just really good. I think there's a natural tendency from institutions to try to reject it first. And that actually for me is like a great sign.
Again, if you look at the history of art, that's been the trend when new generations question the school and the establishment, they're going to get rejected. That actually was what impressionism was at the time. It was a rejected art and they formed the Salon of the Rejected to show their own work and eventually that rejected became the norm. And I think that cycle keeps on repeating where the establishment that has been dominating the art culture sets what the principles and the standards are and then new generations question that and make things in ways that are not supposed to be made and then they kind of effectively end up winning and then you do it again and again and again. And so I think we're at a point where you can make incredibly compelling music and content and videos and films and ads and whatever you want. It's how we want to treat it that will matter the most and I think it's probably institutions already adapting to that.
Said, which is when you're really pushing the boundaries of the status quo, the first inning is rejection. Which is a great sign. Okay, question. I want to shift to the future. You are uniquely positioned to see what the world could look like as somebody who's on the bleeding edge of this. So, as these models move towards real-time generation where video isn't rendered or edited, but fully generated live, what does that change? And I'll give just some examples of what went through my head, but like as we think about the future of gaming or entertainment or even advertising, can you just like stretch our thinking first of all to like what you see as like kind of wild out there possibilities that happen when there is no editing, there's no clean up. It's just in real time. All of us could be viewing something totally different and unique.
Yeah, it's I mean it's hard to predict because you have to like be there to see it and I think a lot of it will be shaped by people more than any other company. So it's less about us and anyone else saying like here are the four or five things. You need to think about it as a general purpose medium. Like the many use cases of the camera were not driven by one person. It was driven by decades of experimentation and tinkering and putting things in different places. And I think with real time we're going to we're already experiencing a version of the same which of course you can think about the obvious ones like gaming is kind of like around the corner you can create very interactive interesting playable experiences with real-time models. Then you can think about tutoring and like live experiences where you can have a version of this right. Imagine I was actually I did an example of this where I was chatting with like Socrates and like Aristotle and you can just like discuss things and you're immersed in a conversation that you just couldn't do before and it feels very natural because you're both seeing each other. You're kind of reading like the way I'm saying things. You can read the way they're saying things. You can think about that. You can think about like companions there like there's many opportunities for people to have someone on their side when they're doing things they haven't done before when they're alone. And there's many interesting use cases that I think we're just like starting to like grasp and begin to experiment with again because it's early these models specifically in real time. We have a model that we released a couple months ago called Characters that allows you to run a conversational agent and a model in that domain. It's only a couple weeks old. And so it's very early beginnings of that new real-time world.
Give me like examples like try to bring this like into the reality. Is this like a game that I can play in real time that is imagined in real time? You just sort of said like discussions with Socrates and Aristotle as a method of learning. Give us some like examples to try to make this as real as possible.
I mean you're kind of like heading directionally like into right like a set of examples and maybe going back to his idea of like mental models and what needs to change. Like when we say a game like there's rules and systems of what a game is specifically when you think about a real tool or like a digital game. But like there's opportunities for us to create a new type of experience that I call nonlinear. Nonlinear meaning like there's no beginning and end and you can take it as an open-ended world and the only shape and form we have to do that in a dead form is a video game. But the moment you open it and have an open-ended world, you can have a nonlinear experience that doesn't and might not feel like a game, right? So, it's like if I spend six hours a day chatting with a fantastical character that's like driving my imagination in interesting places and it's helping me think, what is that? Is that a game? Is that a film? Is that a phone call? Is that an interaction? My position is like, well, the problem is not what to name it. The thing here is that we need to understand that experience as a net new thing. We've never been able to do such a thing in the first place. So, by labeling a game, I'm actually like underestimating, you know, we might think about it like differently just because we're putting it the mental models and the rules of what a game should be like. But if you want to spend six hours chatting with your favorite philosopher and you find it entertaining and you find it the most value you've gotten from a media experience in a while that's amazing. What name do we put to it? I don't know. We'll figure it out.
