Inside Moonvalley’s Marey: Building AI Tools Hollywood Filmmakers Actually Want

0:00For the first time ever, you can democratize studio level films. What that enables is that for the first time ever, you can make things that right now you can only make it a studio because they're so expensive. It like fundamentally changes the entire industry. Welcome to the noise. So this episode is being a little bit different. We had a chance to sit down with Bryn Moser, who is the founder of Asteria and Moon Valley, which we have talked about quite a few times here on the channel. And so we covered a lot. Moon Valley is building out their sort of ethical AI model. Asteria is the kind of company that's harnessing AI power and the production company arm of this.

0:32It was a really involved conversation. And so we're splitting it up into two parts. In part one, we're going to dive into how Bryn has his vision and how he executes on that vision as a leader at Asterio. And then part two. Yeah, in part two, we're going to talk about Uncanny Valley, the film that they're producing, directed by Natasha Lyonne. And so we've talked about that in the past on the show, but we get into a lot of interesting details about how they're thinking of using AI on the film and just sort of the broader context of how, especially budget wise. That was a big common theme of just how they're able to do a film that traditionally would be in the upper 70 millions for in the 10 to 12 million dollar budget.

1:06Let's jump into part one with our interview with Bryn Moser. Absolutely fascinating. Give it a good listen.

1:15Bryn. Yes. Nice to see you. Great to see you, too. Where are we right now? We are in the basement speakeasy of the old Max Sennett Studios, which is a silent movie era soundstage. I think the second oldest soundstage still operating in Hollywood from 1916. That's crazy. So this was a speakeasy at one point, prohibition era. And now it's been, you know, gone through a lot. And now it's our little bar at the company. We do events and things like that. Yeah. Yeah. I realized I have been here before to an event in this bar. it's always like a um one of those litmus tests when people say oh yeah i've been there before because i think there were really crazy raves that happened here like 10 years ago so i can barely remember it man on our team is always like i i saw a picture on the geo located from like 10 years i don't remember and i'm like you're okay you're a dog like i know the people who are partying but um and this was a charlie chaplin's studio so it was it was built by this incredible silent movie sort of pioneer named Mabel Normand.

2:12And Mabel started as an actress, and then she became a writer and a director and then a studio owner. She directed 11 Charlie Chaplin movies, many of them here, and Fatty Arbuckle and everybody else. And her husband was Max Ennett, who was a pioneer of scenic backdrops, which you see upstairs. I can't think of a more appropriate place for you to set up shop. This is so cool. Yeah. Thank you. I mean, it was very intentional. Like I wanted to be in a space where even at a subconscious level, you're reminded of Hollywood's history, especially how it relates to technology. So this was a studio when it was first built, there was no roof because there was no electricity.

2:52So it was all sunlight. I saw that in the model. Yeah, it's wild. Yeah, you know, and then it went through silent film, the sound to color to digital, and now into AI. And I think that our story of Hollywood is always this kind of like, technology comes, changes the industry, and you look back on it, it feels natural, it feels like a normal progression. But when you're inside it, it can oftentimes feel scary. So we're at, you know, arguably the biggest leap in our generation of what's happening with AI. And there is a lot of hesitation for good reason. There's a lot of things that we should be worried about, but there's also a lot of things that I think are pretty exciting that can unlock opportunities for filmmakers.

3:28Yeah. Well said. Good leaping point. Let's talk about Moon Valley. Sure. We met the Moon Valley guys. Initially, I started out thinking about with Asteria, again, the history of Hollywood and what's happened in the past. When you've had these moments of tech change, what kind of studios do you see start to emerge? And you start to see a pattern, which is like a group of multidisciplinary people come together to solve a common problem, right? It's like Walt Disney and the Imagineers or George Lucas and ILM or Lasseter and Jobs with Pixar. It's like people coming together, not to be like, Hey, I figured this out and I'm going to go make a bunch of films, but rather like, how do we all come together?

4:05And like in ILM's case, make a space movie. Right. And so for me, as I started to see AI popping up with Dolly and Midjourney on, you know, social media or discord or whatever else, it was like, some of those things were quite striking. One was just, okay, you can generate an image, you know, that's like interesting but i've put in the context of film like one of my favorite films is la jette by chris marker which is like a seminal science fiction french black and white film that's just still images and a voiceover so i was like well if you can generate images then you can make la jette and la jette is a film yeah uh and then it's only going to get more interesting from here and then i remember seeing that snapchat face filter where you become a pixar character yeah and i was like has anybody shown this to somebody at pixar who's building workflows like could this be helpful?

4:50I don't know. Maybe not. Yes, whatever. But like that conversation should be happening. And so I started getting excited about just like, what if we could find the people all over the world who are figuring out little pieces of this and going viral and whatever, somebody who could make a video and turn it into an animation or make a video, turn into claymation and bring them together and then bring the best filmmakers in the world to start making stuff with them. So that's how we started thinking about Asteria, which is not too different from what came before. And then through that journey, we met some guys from Moon Valley who were leaving DeepMind.

5:18and they wanted to build a visual intelligence model from the ground up. Not one that came from a self-driving car, came from robotics, but one meant to be from the very start. And those were the guys that built VO1, Mateusz and Meek were brilliant. And Naeem would come together. They'd never made movies before and we weren't planning on making models. And so we came together and combined our company. So we became one company. And then, you know, it's been really exciting because I think we have a team upstairs. You guys met a lot of those people who are helping think about workflows. thinking about the way that we need things inside of Hollywood and as filmmakers.

5:53And I was seeing this, you know, I spent about a year meeting every AI company in the world. And I saw a couple things emerge right away. One is, oh man, the technology is crazy. And it's going to be in every part of the way we make stuff. And we're probably not really ready for it. Two is these guys don't know how to make movies. They're not even thinking about the tools in the right way. And three, most problematic, the real big AI companies, like I actually don't think they care about Hollywood. and if you're in Hollywood and you've been here, especially for the last four years, it's hard time.

6:19And so if you've got these people who don't know what the tool, don't know what we need, don't know how to do it. And then they don't even really like, like it, but they see there's an opportunity. That's not a recipe for like how we actually move this industry forward. Right. And in that case, we're going to let it just happen to us. Like what happened when streaming or Spotify or whatever else. So it was really like, you know, I think that artists should have a chance to be at the table with agency and trying to build this in a way that's going to be helpful for us and not the other way around.

6:47I love that vision. And you mentioned a little bit about Dali image journey early on. Tell me about that aha moment that you had when you were like, this changes the game. Yeah. I mean, I think it really was seeing, you know, it was really seeing Dali for the first time where you could generate images and thinking about, you know, La Chate from Chris Marker, which is just still images and a voiceover and being like, man, if you can do, if you can generate images you can make a film yeah if you can tell a story you still have to be a director you still have to be a writer you still have to have a story to tell but if you have a story to tell and you have something that you want to do and you've got that spark and you're hell bent to do it this was interesting thinking about how images could be and then i knew images would would move to um beyond that so that was like i think on the dolly side on the you know for for the on the animation side it was seeing that snapchat face filter just being like well if you can turn into a pixar character like that then like it might be possible that if you put it in the workflow it could be helpful sure um and i think that what was what was exciting is like okay these things are happening but they're missing context and i thought that context was super important because all these things you're seeing like people are making little ai videos that sometimes people are like oh it's like slop and uh much of it is sure because it's not being made by filmmakers it's sort of like and and you know and then the press writes about like hollywood is dead and whatever else and it's just text prompt to video and it's like those aren't movies that's actually not how we make a movie the yeti selfie videos yeah like i mean at least those are starting to become something uh kind of interesting but especially like at the beginning where it was like okay you generate uh like a lot of fake uh you know whatever star wars trailers or something right like people were like oh my god we're in trouble and it was like that's not a movie yeah like it's not a movie it's it but the pieces of it could be it's an exercise in technology rather than storytelling.

8:39Yeah. But what would happen if you put that person next to a great filmmaker? Right. And that's what I think was missing. And, you know, I think that like the AI companies, you know, built, put their tools like Sora in the hands of a lot of big filmmakers. I talked to all those filmmakers and I think that most of the time they'd be like, okay, I see that something's here, but this is not useful for me to make the movie I'm working on. I can't just text something and get a video and put it in my movie. It's not, doesn't have any of the control, the consistency, all those things. And I think that the AI companies really missed the mark here.

9:14I think that what happened was they don't understand how movies are made. I think they looked at what was going on in music. And if you think about like the best music AI tools that are out there, right? And you just put it in to say like, hey, write a 50 cent song about this speakeasy. It would give you something pretty good. Might trick half the people. But like if you put the best tools and put Christopher Nolan there and said like, okay, now just go make the Odyssey using this tool. Never. He would never be able to do it. Now, if you had the best tools and sat him down and said, could you use this to make your movie a little bit quicker or actually help in some of these things?

9:49100 % yes. Some of the shots. Yeah, sure. And I think that that's really exciting, especially in development, especially in previs, especially in giving materials to studios where you spend so much of this time. a great filmmaker with a great vision actually and using these tools can get closer to the creative than I think that they're seeing at 10 pole movies where you're kind of like waiting for a studio somewhere in Thailand to give you stuff to react to for three weeks later. Like that's what I think is quite exciting. We've known each other for a while. I know you don't think of yourself this way, but you're certainly to me a technology pioneer.

10:22Thank you. You've done quite a bit of state-of-the-art technology innovation in the past. And you mentioned a little bit about how big of a game changer AI is, and this is the technology of our lifetime. Talk to me a little bit about how big this is versus everything else you've seen in the past. Sure. My whole life, I think, has been inspired by the moments where technology can really disrupt the industry that we're in and create new opportunities for filmmakers. So I started in a non-traditional way. I was in the Peace Corps. I was in Haiti after the earthquake. But in all those places, I started to see how people were getting access to cell phones, were getting on social media, and all of a sudden, people who didn't really, weren't represented on a stage being able to have their voices heard, could tweet something, could put something on YouTube.

11:09And that idea of like how technology starts to democratize storytelling has always been inspiring, even when we met and we were working on virtual reality or augmented reality. You know, it's always been a kind of thing that I've been inspired by. I think if you look at the history of that democratization. I think we can all look at like what happened with the music industry where you could make a record on a laptop and you see like the birth of SoundCloud rappers or Jelly Roll or Post Malone or whatever. Like it changes the industry. You don't have to be born of the son of the guy who runs Capitol Records to get into Capitol Records.

11:39That's exciting. You know, what happened for me in my career was that when I went to film school, I never had enough money to be able to afford a camera that looked as good as something on the big screen. But then when I was living in Haiti, the Canon 5D came out, you could put a cinema lens, all of a sudden, I could make stuff that looked as good as other stuff that was happening at a film festival or whatever else. And remember, we got nominated for our first Oscar, shooting something on GoPros and Canon 5Ds and whatever else. And it was like, that was a moment where for me in my career, I built my career because of the Canon 5D.

12:11It allowed me to become a filmmaker. And all of my generation, all of our generation, it was the democratization of indie film and documentary. I can now make documentary films. Big deal. Now we're in this moment with AI where I think for the first time ever, you can democratize studio level films. Wow. It's a big deal. Yeah. Because what that enables is that for the first time ever, you can make things that right now you can only make at a studio because they're so expensive. So if you think about the implications of that, it's pretty exciting. If you're only making something at a studio, it's hard to do at a studio because they're putting so much money into it.

12:44So if something costs like$100 million at a studio, You have so many hands in there making so many changes. Everybody's scared. You can't really like put the thing that you want. A filmmaker has to make a lot of, you know, concessions or you've got to make sure that it plays globally. All these other things. Hard to take a risk on something that's a hundred million dollars. If you can do that with the same directors, the same writers, the same stars and do that for under 20, under 10, it like fundamentally changes the entire industry. And then you have filmmakers who create things like, you know, Bluey or Niza 2 or any of the kind of big things that you've seen breakthrough that have become multi-billion dollar pieces of IP.

13:21But what if a filmmaker could actually own that, right? And not just have it be somebody who's hired by the studio. So that democratization of studio level stuff, I think, is thrilling. And if you think about the implications here, it's very exciting because it makes, you know, indie films better. It makes the higher production quality on those things. But if you think about globally, which is what I'm always thinking about, that's the revolution. Like what happens when you bring these models to Bollywood? What happens when you bring these models to Senegal? What happens when you bring these models to Puerto Rico, Malaysia, Indonesia?

13:51That's what's like, I think, the revolution of this moment. Yeah, you mentioned sort of the gap between the existing AI tools and sort of the challenge for filmmakers to adapt these. So how are you thinking about this with Moon Valley and the model you're building? I think fundamentally, there's a form problem in the way that AI was rolled out. They launched these video products. And you have to remember, again, that video and images wasn't something that the AI companies were really thinking about. Remember, everybody was saying, AI is going to come for the first jobs. It's going to be a grocery checker and a truck driver and whatever else.

14:23And then it came for art first. Like, even the AI guys, if you ask those guys, they're like, yeah, we weren't expecting that. And then, you know, I think what happened was as you're building robotics models or self-driving car models, you have to be able to, like, imagine a world. So, let's say you're driving along with a self-driving car. You need to know if somebody walks out in front or whatever. But you also have to have this imagination of, like, a bunch of animals escape from a zoo and there's a giraffe in front of your car and you've got to know what to do. so like those models were training on any scenario happening in a car and then all of a sudden they were like well i wonder if you could just actually make an image of that and so then all of a sudden you're making images and that's like a just a almost a happy accident but then i think when they see that they're like oh there's a product and we're going to disrupt this and disrupt that but and they release it they put it out there everybody writes about it like hollywood's over and anybody who's really making films is like not useful for how i'm making this film like a there's a copyright problem yeah and b i don't want just a random image that this thing made even if it looks kind of like my movie right that's not what i want to do and i think that there was a also a fundamental misunderstanding about how people make and watch movies i think this idea sometimes people talk about like oh in the future ai you're going to make a customized movie and it's going to be you and your buddies and you're in star wars and whatever else i hear that a lot yeah yeah and it's like by the way that's like sexy at south by southwest stage or in a pitch deck but like that's not how people watch movies yeah it's a shared experience 100 and yeah i think also a passive like a lot of people want that passive yeah turn on the tv watch a story experience of course you want to get in it play a video game and by the way i think customized ai video games where you're like playing grand theft auto but it's in your hometown big business and it's going to happen but that's like you have agency when you're watching a movie you don't have agency so this idea that like oh yeah we're going to make a movie and it can be anything you want and then you watch it on your own and then you what you go into work the next day and say like i watched a movie last night about it's like nobody wants to hear your dream i starred in star wars last night and then an octopus came and this like i i don't know anybody who wants to hear somebody's dream let alone like a robot's dream where you're kind of the central character you're talking about pulp fiction but you you do your exactly um and so i think that like that form factor was wrong And I think that the big AI companies who made the first moves thought that Hollywood wanted this thing.

16:47Hey, make me more Bluey episodes, but I don't want to have the team who's making Bluey. I just want to be able to do it. Like this sort of like idea of a studio boss with a cigar. Like now we don't have to have anybody doing that. I tell you, I work with and I know all the heads of the studios. None of them are like that. They're not really like, how do we replace everybody? They're like, how do we make better films? Right. How do we we have a lot of problems in Hollywood. The problems aren't like creative. You know, we need to have bigger creativity or dream more, have tools that make up whatever.

17:19Like we can do all those things. We have problems of like consolidation. Budgets are shrinking. VFX numbers are going up. Like there's a lot of problems. And I do think that the technology has those opportunities to fix it. But the problem of like, wouldn't it be great if you could just text your way to a new Star Wars movie and then you don't have to pay anybody? It's like, nobody's asking for that. And I think that that's where we started to think about context. And so like, if you look at the underlying technology of what is happening, it's powerful, right? And really, it's a combination of a bunch of stuff.

17:50Like one is like, NVIDIA chips are just really great at rendering stuff in a new way. And there's a chips thing that's happening. There's just how fast networks are that's happening. Being able to render in real time stuff is like, it's not necessarily an AI thing. It's just a technological change that's happening. But if you're able to also layer in AI where you're doing all of that rendering in a very particular Laura or fine tune, and then you're able to get information from other parts of the production, like then that becomes something quite interesting. But none of that's really in the form factor of text to video alone.

18:25Yeah. So how are you thinking? Mari is the model, right? Yeah, Mari. Oh, Mari, okay. Yeah. So what is that going to look like? How is the input? How is that going to be different from what's out there? So I think, first of all, what we tried to do from the beginning was think about how do you build AI in a way that's good for our industry? So if you think about two things, one is this is a big leap forward. And if we know anything about history, which we do, it's inevitable. It'll be everywhere. I really believe it's going to be in every part of your workflow in the next couple years. in the same way if you wanted to make a movie without a computer, it would be hard to do.

18:58You could do it, it'd be hard to do. Or without the internet. Yeah, yeah, sure, you could do it. But like, and I hope people still make their movies on film, whatever, but it'll be in all these parts of it. And so if those things are true, it's a leap forward and it's inevitable, then how do we really start thinking about building real workflows so that's good for us as filmmakers, that can allow the things that we get excited about, which is a small team to do more, or have a creator who's closer, a director who's closer to the actual final product when it comes to animation or VFX, or an animator whose style is so unique that they normally couldn't go to a big animation company because 100 animators couldn't copy that same style.

19:32And now you can do that with AI. And so fundamentally, what the Moon Valley guys have been building with Mary is a model that is built for the way that Hollywood needs it, right? So first of all, the model's clean. There's not everything that's in there. We can show it. We've got permission. We paid for it, whatever it is, all those other things that you don't see in the other models. So it's usable in that sense. It's usable, I think, from a legal perspective because the studios can say, OK. And it's usable as an ethical thing for filmmakers who say, yeah, OK, this is an issue. And I see you're doing something around that.

20:05And then in the workflows, it's like, how do you think about plugging that model into the things that we are doing? So creating backgrounds, in-betweening, a lot of development, storyboard, the stuff that you're seeing. Rotoscoping. Rotoscoping is a big one. Compositing. Yeah, compositing. And you're seeing the controls that will be releasing the model imminently. which is why nobody slept. But you're going to see a lot of controls that we're quite excited about, even around things like camera and camera movement in post. That I think is quite exciting that we haven't seen. And I don't think we even fully understand the implication of things like what would happen if you could either slightly or dramatically move the camera in post-production, not just on generated stuff, but on anything.

20:47Very interesting. Something that was already practically shot. So what does that mean for television? And maybe it helps and we don't have to do as many reshoots or pickups or something like that. I don't think we've seen it. We'll see it in the next month of kind of how you're practically starting to use that. Because some of this stuff comes from white papers, right? Like there'll be somebody in China who figures out this thing, writes a white paper, puts it out there. It's kind of the amazing thing about AI is that everything is open like this. And it's just being built up. Like the road is literally being paved as you're driving on it.

21:16So it's theoretical. Well, you could move the camera and here's how you would do it. but it's the context of what would that mean to a television show or what would that mean to a movie um crude analogy so if the if the canon 5d is a camera made by a japanese company for photography and then you happen to use in haiti get an oscar nomination like you know it's like a square thing in a round hole right whereas the sony venice venice 2 specifically made by asc here you know by all of the high-end cinematographers and now that thing is ready for some real films right is it that yeah i think it's i think it's that i think it's when we really set up thinking about building this thing i've thought a lot about pixar and the pixar story because i think that the technology at pixar that was so transformational was render man right like their their their their software basically yeah and and if you if you really just kind of look at what it did it was arguably the first time an animator an artist could have an interface with a computer so before that you had like computer engineers being like oh you can make these little robots move and whatever else and if you look at the things that they were making at pixar before render man it's like a robot whatever falling down the stairs or something and then if you look at what happened after render man and then you had artists being able to use that which is like just two years later it's like pixar yeah look so it's like it's literally story you know and it goes back to walt disney being like yeah animation is cool but can you make people cry we know how to make people laugh or whatever else.

22:46And so you were seeing what happened when an artist could drive the technology and it wasn't happening before. We wanted to build the render man for AI. Like how do you build the artist interface so that artists are able to actually work in AI rather than it just be computer engineers or whoever writes the white paper about it or something. And that's a, I think that's the most exciting thing. And I don't think we've really seen it yet. Like, I don't think we've seen, you know, the filmmakers we're starting to work with and we talk about natasha and uncanny valley you know some of the best filmmakers in the world the stuff we're working on behind the scenes is like thrilling but we haven't seen it yet out in the world we haven't seen what happens when a when a when a great filmmaker has a backing band like all the people upstairs yeah and that's like the unlock yeah that's exciting that's next yeah yeah we're really excited about it all right and so yeah that is it for part one of our interview with brin uh so check out part two we're going to go into more detail about the making of uncanny Valley, directed by Natasha Lyonne, and the budgets and thoughts behind that with using AI.

23:45Yeah. And as you saw, and what we told you earlier, absolutely fascinating mind. Bryn's vision is what I think is really unconventional in the world of AI. And we had a great time. In order for us to bring you more interviews, we'd love for you to engage a little bit and give us some comments, some likes, either on Spotify, YouTube, or Apple podcasts. This really helps us know which direction we're going and if that's the one that you'll have the most value in. Thanks again, everyone. We'll catch you in the next episode.