Exploring Filmmaking with Cuebric's AI: Insights from Pinar Seyhan Demirdag - Ep. 214

0:10Hello, and welcome to the NVIDIA AI Podcast. I'm your host, Noah Kravitz. Generative AI is changing the way content of all types is created, and film and video are no exception. A Gen AI-powered application called Kubrick aims to revolutionize filmmaking and content creation, offering tools for crafting immersive, two-and-a-half-dimensional cinematic environments. We'll get to what that means in a second. Kubrick's technology empowers visionary filmmakers and content creators to bring their ideas to life quickly and effectively. As part of our discussion today, we're excited to highlight Kubrick's participation in NVIDIA's upcoming GPU Technology Conference, aka GTC.

0:50GTC will showcase their innovative AI solutions among the world's leading tech innovators. GTC runs from March 17th through the 21st in San Jose, California, and it's a fantastic opportunity for you, our listeners, to dive deep into the latest advancements in AI, graphics, and more. If you're interested in attending GTC and exploring the cutting-edge technologies shaping our future, we have a special offer. Use the discount code AIPODCAST when you register for the conference at nvidia.com slash GTC, and you'll receive 20 % off your registration. That's AIPODCAST. Don't miss out on this chance to connect with industry leaders and innovators like Kubrick.

1:32As you've probably guessed by now, my guest today is here to tell us all about Kubrick. Pinar Demedra is the co-founder and CEO of Kubrick, and we're delighted to have her on the podcast. Pinar, welcome and thanks so much for joining us. Thank you, Noah. It's a delight. So let's start with the basics. Can you tell the audience about Kubrick, what it is and what it's doing and aims to do for filmmakers and content creators? Absolutely. Absolutely. Kubrick was conceived in a film production stage at XR Studios Hollywood during a visit by me and the other co-founder, Gary, where we visited a virtual production stage.

2:11You know, these gigantic LED stages that are replacing green screens. Ultimately, it's a problem-solution story where our friends who were the owners of the stage were telling us how costly it is to create environments for quality content from Hollywood to medium-sized budget films to everyday creators, that they're all using game engines in order for the camera to track and render in real time. And since we were already in generative AI with our AI production company earlier than Kubrick, we said as a joke, well, who's putting our tech generative AI into your handsome hardware? And they said, no one.

2:52So we always say as two lollipops that never developed a tool, a software. So we didn't know the hurdles and the hardship of it. So naively, we were like, yeah, we can put something together. and then we gift it to you you know you can make cool stuff together with it so we gift it to them we made it they started making environments um from 2d to 2 and a half d for those that are not familiar with technical terms in short kubrick is created to generate an environment with generative ai you can also upload your own environments and very quickly divided into different layers, separate a flat image into different layers for us to have a higher dimensions than just two dimensions, like 2.5D, 2.75D, basically a more affordable and budget saving option for 3D when it comes to environment generation.

3:50So there's a couple things in there I want to ask you to expand on a little bit before we go forward. 2D, 2.5D, 2.75d 3d what what do those mean what's the difference i think i understand two and 3d but but in between i'm a little bit lost yes uh absolutely as you can understand in a three dimensional game you have a gamer and the gamer can walk in any axis on a z depth on a x depth the gamer can go left right and into the depths right left right up down and into exactly yeah whereas a 2d image is basically every single image in your iphone in your photo photo library that's what 2d is right well two and a half d and 2.75 d is everything in between right it's your capacity to take a two-dimensional image rotoscope it meaning another word for rotoscoping technically is segmenting.

4:50Another simple word is parsing the flat image into layers. It's your capacity to identify a house in a flat image and take the house as a separate layer, then fill in the hole that that house creates in the mountain, automatically fill it in and make a new layer of the mountain, and then have the sky layer as a separate layer as well. So Kubrick is born in a film studio with the necessity to streamline the production of dimensionalized still images because there was no one-stop solution for that. The filmmakers were using several different tools and none of them involved AI. So it was very time consuming.

5:33A lot of repetitive, tedious tasks were involved. But as we started entering this beautiful journey to solve a specific problem, we realized that there's also no generative AI professional tool for content creators that lets you work with AI in layers. There's a lot of incredible AI tools out there. I grew up with them. I'm in generative AI since December 2017. But we realized that there is not really one tool that has all the most essential professional creative AI tools that are ready to be entered in professional workflow pipelines. And then out of solving a very specific problem, which is environments are costly.

6:16Here is a cheaper yet quality option to grow the mission into how about we develop a platform where, you know, if you know how to draw in real life, you can draw very well in a virtual environment too. So actually there's no real difference between in real life pen and a digital pen. At the end of the day, the pen doesn't do nothing. Your capacity to utilize the pen is doing something. But that changes with generative AI. So we were very inspired with the idea of why don't we develop a platform where we can let everyone develop their creative genius, not those that were, let's say, born under two artists, parents where they knew how to draw.

7:01So how about we can extract the incredible out of everyone, out of every person that seeks to be creative. So right now, Kubrick is developing tools for the ultimate professional workflows. And whatever we discover there, whatever we learn there, we also give it to the masses. So it's difficult to talk about a visual medium on an audio-only podcast, but we're going to try it anyway. And obviously, listeners, you can do it as you listen, or perhaps when you're done listening, can go to kubrick.com, C-U-E-B-R-I-C.com. And is the app named after Stanley Kubrick? Of course. You can see, wait, look at my library.

7:47Sorry, listeners. Now I can see my library. The whole section of library is Stanley Kubrick. I can attest, Pinar's pointing right now to all the Kubrick, all the Stanley Kubrick books behind her on the shelves in full 3D. When I feel down or when I feel sick, I read books about his life because I know every single part of it. Oh, that's fantastic. Very cool. What's your favorite Kubrick movie? Can you pick one? No, you're shaking your head. Eyes Wide Shut, it has to be that. Yeah. Okay. Eyes Wide Shut. If you haven't seen it, go see it and then come back, resume the podcast, check out Kubrick.

8:26But let's talk about Kubrick. You've got a great tagline called concept to camera. Can you talk a little bit about what that means and sort of conceptually what you're trying to do, but then from the user standpoint, how does that work? I would like to start with a very large vision and then I will trinkle it down to what it means for today. Great. concept to camera have you i hope you guys have seen avengers endgame if not go to youtube just type avengers endgame tony stark inventing time travel scene where the main lead character basically has this holographic dashboard called jarvis where the lead character basically comes up with the idea of seeks the curiosity of how the time travel works and where there's a tool which is called Jarvis, which at the end of the day is voice to 3D, voice to holographic 3D.

9:23And together with the two, the lead character invents time travel. We always refer to Kubrick as the Enes Dost version of Jarvis. So that's actually, imagine like the guy has an idea in his head. He doesn't even have an idea. He has a seed of curiosity, how would time travel works. And the tool helps him extract and visualize and like visualize all the quantum possibilities of how can he get from a concept into a manifestation of his concept in a screen, in a holographic reality, whatnot. not so if we were to have that as our north star you know as a creator be it a bedroom creator be it a corporate creative that's visualizing a presentation be it in the creative department of a tech company whatever whomever has an idea to create anything they need two things one a talent an actor, a present, presentator to an environment.

10:32Okay. So the actor is here. You don't have to manifest them. You're just, they're just there. Right. But you need some form of a vision and, and production for the environment where that talent exists. Right. But there's a lot of limitations and there's a lot of tedious tasks that are placed until from an idea that was once in your head that ends up on screen. And it is our responsibility as Kubrick team to relentlessly research that. What are the most tedious and repetitive parts? What are the boring parts of you conceiving something, get it approved and parse it through, repeat it, clean it, reiterate it.

11:12So we constantly are in the research of identifying the tedious parts. So then we can optimize them with several different AIs, including generative AIs. and have you be left with the most important part, which is the creative spark. Right. Anything that can get my vision directly from my brain into a medium that other people can understand, I'm all for. Yes, exactly. And so how does it work? We read your mind. Perfect. We're done. Thanks for coming on the show. Remember to sign up for GTC. We run on telepathy. Yes. One day, Noah. one day I will get there. You've got my email address. Let me know ahead of time.

11:57A lot of Gen.ai image creation systems work on text prompts. And you mentioned before about the analogy of Kubrick being kind of the MS-DOS version of Jarvis, so to speak. Is it a prompt-based system? Yeah, absolutely. Prompt-based is one of the features. We have 10 plus different features. Okay. Text to image is definitely one of our main features. We have four proprietary models that are cinematic. And then we have one graphic model that yields an animation comic book type of outcomes. But you can also upload your own image in Kubrick. And then the most fun part that comes along, which is our AI-based segmentation methods.

12:43Okay. AI people say segmentation, film people say rotoscoping, everyday people say separating a flat image into layers. They all account to the same thing. So we have object-based segmentation, object-based rotoscoping, which is basically if you have, let's say, a girl like me or, I don't know, a mountain, a house, a tree, some very visible objects, our object-based segmentation works very well. You basically hover on an image, you click, and then it's a, whatever you clicked on as an object, it's a new layer. When you're talking about creating layers, is the idea that that's what you need, this is my own ignorance in the world of 3D, is that what you need to be able to view the object from different perspectives?

13:29Or what's the purpose of... Yeah, sure. Thank you so much. I should have, yeah, talked about that earlier. You are on a film stage and it's made out of a gigantic LED stage and you are an actor and you have a camera. So when you walk right, if the environment doesn't render right with you, it looks like you're on a fake set. Right. It looks bad. It's not it's not believable. Right. So a solution would be to create an entire game as your background. So then wherever you walk and talk, the environment real time renders with you. And that's an incredibly expensive solution. Sure. Like$1 ,500 ,000 plus starting type of expensive solution.

14:13So a plan B for that would be to dimensionalize a flat image into a layered image. So then it gives you the chance to move even a little bit. So for example, as you know, when you are walking in a forest, let's say, the speed of a mountain at the very back moves while you walk is a different speed than the tree that's closest to your face. The tree moves very fast. You just passed it. You just made a step and it's no longer there. It's the same mathematic when it comes to making separation of layers, flat layers into different layers. So then when the actor is walking right, the tree that's at the very back of the actor moves in a different speed than the mountain at the very back.

15:01It moves very slowly. That's the very reason why we are dimensionalizing images. It's the same with animation. It's the same with game development. I would argue that half of game development is made on phones. and phone games do not require very hefty game engine renderings. They're perfectly well, like most of the phone games are made of 2.5D environments. Got it. And so Kubrick is offering a faster, cheaper, easier way to render realistic environments than leveraging a game engine. Exactly. Right. Terrific. Exactly. That is our production part, but we're also a very good concepting tool. How does it work?

15:49You again, generate your image in Kubrick. You can upload your images and you can say, hold, you just uploaded a girl sitting on a chair. Instead of the girl, I want the same chair. I want the same everything. But where the girl is sitting, I want a dinosaur. So you can have that control or you can be like, okay, now, you know what? I would like to modify the pixels right over her shoulder. Just select a tiny area over her shoulder and write a parrot. So you can make a parrot sit on her shirt. Right. So it's a very good tool that allows you to work with AI in layers and that enables you to have fine control over the modification of pixels during your ideation period.

16:41One of the things that I've found in a lot of the work I've done with Gen AI, when I'm not podcasting, I do a lot of writing work. And so one of the things I've found Gen AI great for in my own work is exactly that. It's brainstorming and sort of being able to very quickly ideate on, you know, well, what if instead of talking about, to use your example, instead of talking about a girl, we're talking about a parrot or a dinosaur with a parrot on its shoulder. And it's easy to kind of just step through these different ideas and brainstorm it out and then say, oh, no, let's go back to the other thing.

17:15How are our filmmakers and video content creators using Kubrick now? And maybe even a better question is how do you how do you see tools like Kubrick? Well, they're already changing, but continuing to change the way film and video production is handled going forward. Well, that's a very good question. This is such a large topic that I would like to give a metaphor or another example of something that gigantic left an energetic mark in the development of humanity. And that was going from horses to cars. It's a paradigm shift. It's not just like faster horses, but an entirely new behavior change towards transportation.

18:08It's the same way with content creation, be it film, be it social media, no matter what, there will always be screens and there will always be more demand for content in screens. And we always say that there has never been this much of desire and demand for that much extraordinary content. Like, will, will businessmen, will like lawyers stop being lawyers and become content creators. One way or the other, we have to satisfy this demand while keeping quality. So the most different, I'm actually also the generative AI instructor, main generative AI instructor of LinkedIn. I had my first course that was called, What is generative AI?

18:51Oh, terrific. Okay. Yeah. And my second course that came out this week, that is about AI aptitude. Until that course, I didn't know what the word aptitude meant. I'm that serious. Apparently, basically, it means your capacity to discern what tasks should be given to AI and what tasks should be handled by humans. Which is so critical. Yeah, exactly. I predict that there will be challenges unless we truly understand who we are today, what humanity means, what is our difference, what is our superiority, what makes us incredible sublime beings. unless we understand that today. I know that there will be some rocky periods that we will be facing tomorrow, which is understandable.

19:37But it will be a fun ride where, like we just have to, like we're going to rethink our entire approach to self-expression. We will no longer be only expressing ourselves within the limits of our budget or within the limits of our technical capacity. or within our capacity for our project to be approved by a conglomerate or by a studio, we will all become our own bosses and our own studios and our own masters. It will be like everybody's, not everybody, but a lot of people are afraid of AI, and I understand that. But actually, if somebody needs to be afraid, it will be conglomerates because generative AI at the end of the day is human empowerment tool.

20:27I'm speaking with Pinar Demedra. Pinar is the co-founder and CEO of Kubrick. And in addition to some pretty wide-ranging, oh, what's the word? Not proclamations, but talking in big terms about the future of humanity and self-expression, Kubrick is a generative AI-powered, or AI-powered, I should say, generative AI, one of the models being used, production tool for filmmakers and video content creators. As you were talking about what AI means and might mean to self-expression and the ability for one person to become sort of the equivalent of a content production studio, I kept thinking about, this might sound weird, I kept thinking about Vancouver.

21:15my you know i've been an avid uh consumer of tv and film and video my whole life and i've played around with the video editing and making youtube videos but i'm not an insider but i i've kept hearing over the years that a lot of film a lot of movies and tv shows set in different cities actually were all filmed in vancouver canada and there's something about vancouver that lends itself to production work that can be made to look like i don't know new york city or wherever else And so I bring this up because I was thinking about Kubrick in the context of, are people using it more likely to use it, do you think, going forward to replicate kind of known environments that, you know, it's easier and cheaper and faster for me to use an AI tool to make it look like I'm in Vancouver than it would be to actually travel there and create things?

22:09or do you think we're headed towards more kind of, oh, I don't know, abstract and kind of unknown visuals? Whereas instead of making it look like I'm in Vancouver, I'm going to make it look like I'm in this setting that you've never seen before because it came directly from my brain and incorporates elements of places on earth I've been, sci-fi novels I've read, maybe anime that I've watched. And I've kind of now have these tools that can let me sort of mash these things up and play with them from the confines of my laptop and create an entirely novel environment in which I'll place my talent to do the acting.

22:52Where do you think these things are headed or have you seen any examples of people using Kubrick in some of these ways? Yeah, it's been used by several different directors. Right now it is used on a very beautiful project called Europa, Europa. Okay. that is produced by ETC, Entertainment Technology Center of USC, University of Southern California. Okay. It's currently being filmed at Sony Stage by an award-winning director. I cannot say much about it, but for those that know, Europa is the name of a planet that has the most topographic similarities to Earth. Okay. It's a common knowledge. So obviously it's a science fiction.

23:37and yes, it is used very well in a science fiction setting. But to answer your question, I think it's more like a supply and demand question. Kevin O 'Leary has this really viral sentence. He said, when I was growing up, if people were to ask me, what shall I study? I would say, be a coder. But today when they ask me, I say, be a content creator. In the same way, I believe the demand for humanity to seek more of mind, theory of the mind type of content is raising because we have started asking questions to ourselves that we were not asking 10 years before. Right. More of what is the meaning of consciousness?

24:21What does it mean? What does awareness mean? Can you define God? What is the power of nature? Parallel universes, simulation theory. these are very prominent topics of conversation in the contemporary culture. So obviously, these are very hard topics to visualize. You can't measure God of like, you know, what is the height? What is the width? So, of course, it's got to be 3D, not 2.5D, though, right? I can debate you for it. All right. Fair enough. Another hour about that. But yeah, so yes, obviously generative AI-based, imagination-expanding tools like Kubrick are great for theory of the mind, things, pictures.

25:08Do you have any advice, whether it's sort of large-scale advice like you were just talking about, or kind of more ground-level tips and tricks for aspiring filmmakers and content creators who want to start leveraging generative AI in their work? Yes, absolutely. I always give the weirdest answer to this question. I expect nothing less at this point. And I mean that in the best way. Yeah, thank you. I would say seek to understand what subconsciousness means. It's the weirdest answer. Please let me explain why. Please. When I was first introduced to generative AI, end 2017, beginning 2018, it was also the same time that I discovered what subconsciousness, conscious and unconscious means.

25:58And I quickly could draw a parallel between the mind of AI, which is a data set and an algorithm, giving you an outcome. Right. It's simple, simple 101 AI data set algorithm gives you an outcome. You want a different outcome. You can't bicker on the outcome. You have to change the data set and or the algorithm. Right. Actually, the human mind works the same way, I understood, that our subconscious is actually our data set. Our belief system is our algorithm and our personality, which is the belief in ourself, is the outcome. So if you want a different outcome in your life, if you want your movie to be approved, if you want to have more self-confidence, if you want to be independent, if you want to be wealthy, whatever you want, be it success, professional, personal, you first have to change either your data set or your outcome.

26:51AI, of course, will come as an empowering tool. But if you refuse, if you refuse to be empowered, if you refuse to be a better version of yourself, no tool can help you. Be it AI, be it spoon, be it laser beams, nothing will help you. Only you can help yourself. I'm tempted to make a joke about how listeners came for the latest on AI and they're leaving life advice. And then I realized, you know, we're talking about filmmaking and content creation. So of course, this is where we're headed because what's the point of filmmaking and content creation like this if not to explore and express the human condition?

27:28So it actually makes a lot of sense to me. Looking ahead, Kubrick's going to be at GTC, as we mentioned. Any upcoming features or announcements that you can talk to or tease or anything you mentioned, the Europa project, anything else around how Kubrick's being used that you might want to talk about in closing here. Sure, thank you. Yeah, we recently launched an important integration in Disguise. Disguise is a media server. Media servers are in short, basically the translator between your content and the LED stages. An LED stage is that I'm picturing like a green screen, but instead of a green screen, it's an LED wall.

28:09Yeah, a gigantic LED wall, a curved gigantic LED wall. These are called volume stages. Another technical term is virtual production stages. Or you can also refer to it as LED stages. They all mean the same thing. Right, okay. So Disguise is basically the media server, the tool that enables you to create real-time effects and real-time renderings in that gigantic engine. Okay. And Las Vegas Faire is powered by Disguise. Disguise was the media server of Batman, bullet train joker so you know then they're no novel i mean they're novelty of course but you see what i mean they're the best of the best around around the block for a few years sure so kubrick and disguise announced a really important integration a month ago we launched it in la so with this integration you start a two-dimensional image in kubrick like basically a flat image from your phone that let's say you made a environment like you made a picture of your backyard or you generated in kubrick sure you very quickly turn a flat image into a layered image in kubrick like i explained during our podcast with our in painting generation ai based segmentation tools but then the magic comes when you push the button export to disguise then everything opens as 3d in disguise right already all your two-dimensional layers open as 3d layered because Kubrick can read the meter distance between every object in a flat image.

29:48Okay, right, right. And I believe to this day, we are the only tool that does that. Like not even a human can look at a house, mountain and sky and be like, this house and this mountain is two meters. Yeah, I sure can't. Yeah, me neither. I'm lucky if I get the direction right. But yeah, so this is quite a remarkable achievement. That's fantastic. It's all coming from a flat image, which is quite crazy. So that's what we announced lately. Pinar, for listeners we mentioned before who want to learn more about Kubrick or even try it, they can go to the website, kubrick.com? Absolutely. They can enter in the app directly from app.kubrick.com.

30:31And it's suitable for a bedroom content creator, as you said. They can give it a try. It's fantastic. It's so easy to learn and use that you can, with our YouTube page, Kubrick Official, you will find all the necessary tutorials, but you don't even need that. It's so user-friendly. Great. Well, I think when we're done, I'm going to try to delve into my subconscious a little bit and wait for something to bubble up. And then I'll head over and give it a try because I want something good to work with here. So, Pinar, thank you so much for coming on the podcast. Congratulations with everything so far.

31:08The integration you just announced and what sounds like just, you know, the, if not the very beginning, at least the early innings, as we like to say, of this brave new future of AI augmented, you know, content creation and expression. It's fantastic to think about. And I hope to run into you at GTC. Thank you. I actually, before we close, I want to thank the whole NVIDIA family because we have been welcomed to NVIDIA's Picasso program. And I believe that in the beginning, they only welcomed six companies and Kubrick is one of them. And we are still in our process of self-funding. Kubrick is founded by its co-founders, me and Gary, the other co-founder.