The Genesis of USD

Announcer:

Today on Building the Open Metaverse.

Guido Quaroni:

Yeah, I know. It's tricky. When something is not fully understood, it's really difficult to get together and standardize it. That's why I said, "Okay, geometry. I think we're there. Shading network. Eh, maybe getting there. Rigging is getting a little painful. And now you're going into full simulated system, so I think it would be a while before there is clearly the 90% of the definition. How do you define forces, collision objects, and all those things to really set the condition for a simulated system in a standardized way?

Announcer:

Welcome to building the open metaverse where technology experts discuss how the community is building the open metaverse together. Hosted by Patrick Cozzi from Cesium, and Marc Petit from Epic Games.

Marc Petit:

So hello everybody, and welcome to our show Building the Open Metaverse. The podcast where technologists share their insight on how the community is being the metaverse together. Today I'm with my co-host Patrick Cozzi, CEO of Cesium. Patrick, how are you?

Patrick Cozzi:

Doing great today, Marc. Happy to be here.

Marc Petit:

Happy to be here. We have a very special guest, somebody that if you are in the VFX industry, we've known for years. Our guest today is the father of USD, Guido Quaroni. Welcome to the podcast. We're so happy to have you here.

Guido Quaroni:

Yeah. Thank you. Thank you for having me, Marc.

Marc Petit:

So let's introduce you to the people here. You've been at Pixar for 24 years before joining Adobe a year ago. And I was mentioning that when I Google your name, we find voice actor. Can you tell us a little bit why you are known to the world as a voice actor?

Guido Quaroni:

Yeah. So this was ... A story was about early 2000 when Cars was being developed, in the story. And there was obviously this characters, was Italian characters, Big Ferrari fan. And they start looking for names. So I remember one day I was actually talking with the directors, John Lasseter, about one name we could use. And I suggested, "Well, if you translate “I drive” is Io guido. It means actually guidare is “to drive." So I said, "Guido is not a bad name for a car." So initially it was a little bit like, "Okay, but coincidentally it's your name."

Guido Quaroni:

But at the end they checked, and other people from Italy say, "Oh, Guido's a perfect name for a car." And that was the deal. On top of that, actually I got to work on the actual character itself. I did the shading, I did the hair, there's a hair with the Italian flags. It was kind of cool. It became my little buddy.

Marc Petit:

Pit stop, pit stop. Right?

Guido Quaroni:

Pit stop! Yeah. The lines were not that deep. It was a pretty simple character. Not a long time to memorize them.

Marc Petit:

Well, look. Your contribution to Pixar goes way beyond the voice of Guido in Cars, I think. Can you tell us a little bit your journey there, and your journey to the metaverse?

Guido Quaroni:

Yeah. So I started back in 1997 at Pixar. And at the time was working on Toy Story 2. So I was a TD, at the time, everybody was doing a little bit of everything. So modeling, shading, rigging, effects. Whatever, because the crew were so small. So work on that movie, work on Monsters Inc. And then move for a number of years in R and D, to really improve the shading tools. And that was the thing that I really become passionate about, the 3D paint and look development, shading tool, et cetera.

Guido Quaroni:

So a few years there, and then back to production. When Toy Story 3 started to be developed, I applied to become the VFX supervisor for that movie. And then worked on Monsters University. Pretty much Toy Story and Monsters are the movies I worked really actively on it. And obviously then there's Guido on Cars. About in 2011 in the middle of Monsters University, I took over the R and D group, the engineering team at Pixar, developing Presto, the animation software and a bunch of other stuff.

Guido Quaroni:

I always loved the idea to be a user of the software, and also a part of the design and the development of the software. And sometime when I was in production, I got mad at something, I went in and just start coding a few fixes here and there. So that was the part that I enjoyed the most about being in an environment where you can do both in a way.

Marc Petit:

Yeah. That's given you a pretty unique set of skills right there.

Guido Quaroni:

Yeah. And then again, in beginning of the year, I started at Adobe as the director of engineering for 3D and I know Sebastian from way back, and I wanted to try to even look into even a broader, a wider scope to software that is not just specifically for movie making, but engineering, manufacturing, design, et cetera.

Patrick Cozzi:

So Guido, one of the topics we were hoping to cover today is USD. And we know that, that you launched that when you were at Pixar. And some of my earliest encounters with USD, I've done a lot of work with, with Khronos and the glTF format. And when we learned about USD getting all this traction, even outside of the movie industry. We've met with some folks using USD at the Game Developer Conference, and I'll never forget the quote. They said, "USD is not a format. It is a way of life." So we would love to hear the origin story for USD. How did you do this?

Guido Quaroni:

So that was again about more than 10 years ago now. Pixar had developed some internal ... As part of the animation system developed a seed representation for that system itself. But it was proprietary, and at the same time we were looking at Alembic, we had some other cache file format internally. And then we become really a mess at the studio itself, because we had to go through multiple APIs. You have to go like, "Okay, if I want to read this, I'm going to have to learn Alembic. I have to do that, I have to learn something else." And so then the question was, "Well, we have something interesting here and what do we do about it? Should we actually bring it out there and see if it can become a standard?" Again, it's not just a file format, but it's a scene graph description.

Guido Quaroni:

I think we learned a lesson with RenderMan. RenderMan never opened the RenderMan Shading Language. And today RenderMan is using open shading language. The RenderMan Shading Language died. And what happened is there was a time, there was an open system. It became very similar from a power and flexibility point of view, it just took over. And so in my mind, it was all about if you don't do something here, one day, we're going to throw this thing away. Because something else will take over. So my pitch to, at the time was Ed Catmull was the president. And the executives was, it actually is a competitive advantage for us to put a system out there that if it takes on, and is a standard, guess what? We really like it, because of the way we work. And so that was the spin. It was really about ...

Guido Quaroni:

And the second aspect that drove me for that was giving an opportunity for the engineer to do something that impacted way beyond the actual 3, 4, 500 users that we have at Pixar. And so to me it was also a retention and a sense of pride. And for the production people, releasing a movie is their moment. And I wanted for our engineers to release something like this, to be showing their work to the world. Also, it felt by going open source, the bar had to raise. Not that we had that standards on everything, but once you got to show it to everyone, you do some extra checks in there.

Guido Quaroni:

So there was all those ... And so we started working toward removing all the dependency from all the other file formats et cetera. And tried to do was only one system for everything that picks up. But as we went through that, designing in a way that could be open sourced. So no crazy dependency, make sure that no copy and paste from code from somewhere, and stuff like that. And then I started, and I remember with Marc was in those conversation early day. Keep throwing a bit, asking the question. What do you guys think? To the studios, and to Autodesk. If we put something out that is Alembic, GLTF, all this file format I see them like the JPEG of the images. And this is more like the Photoshop file. And so it's a much higher level description of the world, and the other ones are more the leaf level.

Guido Quaroni:

And so that was the pitch about, "Hey, what about going beyond the individual asset, but the aggregator of the asset? How you bring them together, and how you do the overrides?" That it wasn't quite written into the existing systems. And as soon as I start seeing ... I actually realized that every studio had something similar, and they all were doing the same stuff. And I said, "You know what, maybe we just want to go first. Do a good job. And if it takes off again, like I said before, we'll be happy about it because we like it."

Marc Petit:

Yeah. There was interesting anecdote is when Alembic got released, everybody realized, "Oh, I wish it would've been my format that open sourced then." But the island guys got quicker.

Guido Quaroni:

Just a little fun an anecdote, the original name ... And we were looking at the extension for the file. Technically when I do something that says layered scene description. But the extension LSD we thought it was a little too trippy, so we decided that we couldn't go in that route. So that's why the universal came, because it feels a little too far, but technically it was supposed to be a layered scene description. But anyway, we couldn't use that.

Patrick Cozzi:

Yeah.

Marc Petit:

Although USD has kind of has a connotation to it. All right, well. I was mentioning, and Patrick and I, through this. It's probably our episode eight, we've heard about USD every single episode. I think people are enjoying the prospect of having interoperability, and USD's bound to play a good role. But we know USD is very declarative in its aesthetics in description. How do you see the role that you can play in the metaverse and how can we evolve there to have more of an execution model, and contribute to that exchange of simulated world? Which is I think the overall ambition in the metaverse, is like let's trade a complete simulated world. So how do you see the roadmap to that? Or the milestones?

Guido Quaroni:

Yeah. The way I see when we think about the metaverse is there're almost two layers. The first layer is how do you define in the clarity way what's there? What is this data? And I don't think we can go in a completely procedural ... I mean, I don't know, maybe we could one day. But the idea would be you describe buildings and things and objects, and you define what's there. And that's what USD is good for. The tricky part is how we add that layer of a runtime, that almost takes that ... Almost like a starting condition. You have a starting condition and then from that starting condition, you evolve.

Guido Quaroni:

And that's actually not necessarily ... has to be everything part of USD proper in the sense that ... I can imagine actually to having the layer that it's above that, that does the simulation. And it changes the state of the world, necessarily. But then at that point, the clarity world anymore because things are driven by events and things that can happen. So actually, I wanted to avoid a bit the assumption that, "Oh, USD is going to be everything." It's almost like ... Because the risk is that try to make it to try to be too big, to make too many things, and then implode on his own weight. So that's why when I think about a game engine, or when I think about a simulation state, I see it as initial condition, I create the state and I go, and I start. And in time, start going in a more unpredictable way. There is pieces that are predictable pre-computer, and that's why USD is okay with it for cache data, but the more unpredictable ...

Guido Quaroni:

Now there is definitely a conversation about what would look an execution system, or actually an engine that sits on top of it that we use as much as possible. The same schemas and definition of the what's in there. So then the transition become more natural. So the way you define a mesh and the way you define cameras, and all of that. And how do you change those attributes now that are not driven by something that is pre-computer, but is actually driven by dependency or relationship or constraint, and things like that.

Guido Quaroni:

That's where I think the next aspect ... It's almost like when I think about the metaverse, I keep imagining there is a metaverse browser, that it's actually you send links and USD is like the HTML of it, that gives you the ... It's not a web page, it's a web space now. And so then this browser ... And now it's a browser, then on desktop looks like a browser, but in VR. You get in there so that it's more immersive, but conceptually has the same model.

Guido Quaroni:

And so there's almost like a Java script or something on top of it, that gives you that execution of it, and things that can happen based on triggering from events, and things like that. And maybe at some point again, I know maybe I'm going too far up, things that have their own behavior. They don't really have to be everything to be described upfront. So USD can play a role to be, I'll give you the topology and the structure of what's in there, the way you have an HTML page. I'm making a loose knit thing. And then there is something on top of it that lets you play with it.

Guido Quaroni:

Actually the more I think about it in the ... Maybe one day we'll see a browser from Epic, or from other company. Is really about this idea that we will share. There won't be a single metaverse, necessarily. The way we don't have a single webpage to do everything, we have lots of ... But it's very convenient, the idea that you jump around by maybe having a link, the way you jump around pages. And once you're in there, the browser transports you in the right way. Maybe if you have 2D element that become a card in space, or maybe flattened on the screen like the old HTML. But the actual scene that can be described with USD, it's actually what's driving into the browser. And again, on desktop will be flat, on the screen on VR, you are immersed on AR. Maybe is an alpha overlay.

Guido Quaroni:

But that's the model that I'm imagining, is a transitional. And again, USD is the skeletal part ... I mean, I'm not saying that USD has to be, will be, or is the best. But right now seems to be a possible player or something similar like that.

Marc Petit:

Yeah. We had Vladimir Vukićević from Unity discussing with us the other day. And he was advocating the extension of something like glTF with properties and even behaviors that could be potentially standardized. Like something is hot and ... Do you think that approach could, could fit well as a compliment to what we would need USD to be, as the declaration of what the world is like?

Guido Quaroni:

Yeah. I mean, there are a really extension to USD. I think Apple as well has done the one with the behavior. And that will definitely be the first start. I think it's maybe still limited. When you click, you do that, but that would be definitely the first iteration, would be event driven, things that change into the scene, so then by user input. What I'm seeing that is even more also likely is that time goes, and simulation start going. And so at that point, the system ... You give even property like mass and collision, kind of stuff. The system then goes on his own. So it is not much event driven. I mean, there are some of it, but a lot of it may be completely ... Again, then you need the crazy amount of compute and all sorts of other things. But I can imagine that to be ...

Guido Quaroni:

But yes, that the way glTF and ... So thinking about how can we create another layer on top of USD to add those behavior. Maybe it's new schemas and extension. My hope is that it's done in a coordinated way, meaning that not everyone ... Because Apple goes off and maybe does one. And then if somebody does something else ... And again, Apple started that conversation for sure, but it has to be, for example, for simulation. There was a group discussion among some of the vendors, because we want to avoid ... Again, this only works on the iPhone and this only works in Chrome, and that kind of, stating the obvious. That's kind of what we always live without, that issue.

Patrick Cozzi:

So, you just started to answer a question I actually wanted to ask. So I loved your analogy of the PSD versus JPEG, for USD and glTF. And yeah, I wanted to ask if there's anything that you think that the glTF world should be doing to help facilitate that collaboration?

Guido Quaroni:

Again, it's the same with that parallel. glTF I believe ... Unless glTF changes, I believe glTF is serving a very good purpose of being a very efficient, leaf node of the whole world. More than an aggregating system and a tribe. Because the big different with USD, that is a runtime with composition. The runtime does referencing, the override ... It's not just the final desk, there is a lot more. It's the aggregation component. So for glTF, I think it would be great to keep glTF to be the best into that what is serving today. So when you have an asset that has to travel really fast to be presented on the web and all of that, I think it's still a valuable thing to do.

Guido Quaroni:

Now. Somebody may argue, "Yeah, sure. Even new USD can be a leaf node." And that could be, arguably. But I'm more worried about glTF, and the same thing was happening with Alembic, but it didn't went all the way, start to try to become USD. Like, "Well, let's have referencing, let's have override." Because really that one is not just a specification. You need to write a runtime for it. And so, and the that's why my hope also that some pieces of this runtime remain open. I was talking to Nvidia about, "Hey, there is anything in the Omniverse thing that you guys have done can be open source. Because it would be great to have some of those runtime improvement to be consistent." And because if we start having more, that only works there and not up there kind of thing, it gets tricky.

Marc Petit:

Yeah. Or I mean, Michael Kass a couple of weeks ago told us that they've done massive changes to the compositing code in USD to do performance speed up. I agree that it would be fantastic that those can get contributed back to the community.

Marc Petit:

And one conversation we had, I'm curious to have your opinion about, that balance between an open standard which lets people implement it differently. And so allows for software, because you're now a software vendor, right? Allows software vendors to compete on implementation, and it's something that is Khronos is managed anybody can come. It's a relatively quote unquote democratic process, where everybody can become a cross member, has a voice on how the standard will evolve, and then software vendor can compete implementation versus the open source model. Where Pixar, your former employer, remains firmly at the wheel and deciding what makes and what doesn't make, but on the other hand gives you a proper implementation. So how do you look at those two different models? And do you think that USD as an open source project can evolve into being the thing that a lot of people will depend upon?

Guido Quaroni:

Yeah. My hope is actually ... And again, it was part of those conversation about making USD part maybe of a foundation, or being more open there. I really love the idea that if we continue to improve what's in that open source repository. And the reason is because it really makes sure that ... Because it's tricky. When you implement open standards, then there's always who implemented better than other, depending if the hardware you want to push. I don't know, OpenCL was a good example and some other ones were like, "Yeah, the standard is there, but in some platform doesn't work and another doesn't." And so then, if I think about on the other side from the who's experiencing it, the more is guaranteed to be there is a certain amount.

Guido Quaroni:

What I think would be great to keep it open source is the part that resolve what it looks like. The actual ground truth of the data. So a sphere looks the same sphere. Now, if I got the most amazing system to display and interact with it, that totally should be different. What I worry about at the moment that there is something that is now start to resolve differently. And now, "Oh, it kind of does." Yeah, sure. The spec is there, but it's quite not the same thing because I took a shortcut. That's why when I was talking to Nvidia to make sure that if you're introducing things, then now you're needing special metadata, or other things that are just specific. Suddenly the resolution of the data is not consistent. So where things are placed, to oversimplify it, right? If a sphere is here is supposed to be here, not here on this platform, and there in that platform for whatever reason.

Guido Quaroni:

But I agree. I think if we don't get to that openness also the open source side, then they will be automatically driven to start doing their own. At the same time, I'm at Adobe now. And every time I say, "Hey, we should do this and open source it." Clearly there's a tension there. Because they want to keep a competitive advantage. When it comes to USD the place Adobe wants to place itself is really to make the best tool to create. We call ourselves the Creatorverse. We're like right now we want be best authoring tool, creation tool, to sit on top of standards, possibly. So creating awesome USD worlds, that would be our hope. But again, I wouldn't be surprised that at some point there would be a like, "Oh, shoot. We have this thing that only in our tools you can deal with it." But right now I'm just trying to, I'm a firm believer.

Marc Petit:

Good. Yeah. It's fantastic. I think, can talk a bit more about Adobe. You guys did some recent demos and there seems to be a very nice portion, or tendency towards procedural in everything you guys do. So how do you reconcile that appetite for USD and that level of procedural content, procedural materials and all of those things?

Guido Quaroni:

Yeah. The idea would be to describe the procedure with USD. And when you get into the billions of things, I don't even know at that point that you need an actual description for it. Just as an example, hair. I was involved with making the hair system, me doing the hair system. I developed a system that was used for a number of years, I think until Incredibles 2, like 10, 15 years, a long time. But anyway, the system was all about, you define some low level detail, like a scalp, a guide, a taxi driving, all that, and then you amplify the data. Now, the way you describe the procedure was the declarative, the way you actually visualize it is totally procedural.

Guido Quaroni:

And it's actually that the render time is in the GPU. You don't even visualize it in a what's in there, you're just you're shown in the view port. So the way you're imagining is to USD being a good place to describe a procedure, like the connected node into graph that describe the operation. And maybe there is some standardization of what that nodes does, what's the name? What the parameter? That goes back to my interest in material X, or the way to standardize that. And maybe one day somebody will come up with an interesting standard to spec. How do you take an object and you line 20 of them on a surface, or like a geometry procedural.

Guido Quaroni:

So USD described it, but then that's where actually the game is played in the game engine. And that's what actually, maybe that's where you played the proprietary game. How you actually visualize and you resolve all that complexity. Because USD will not be the right place to put 10 million hair and put them on desk every time. If you have to, you do it, but you really don't want to do it because the increase in complexity is way outpacing potentially in that space. I mean, [inaudible 00:25:08] is a perfect example. You're dealing with super complex problem. When you go into the billions, you're not going to use standard declarative models to describe a mesh. I read some of those research, it was amazing work. So USD at that point is going to let it go. The scene is there, there's a bunch of procedure described, now I'm going to resolve it and data can explode to another level.

Marc Petit:

Yeah. The other aspect about the metaverse, which is a simulation aspect I mean, you can look at the game engine as a bunch of systems. We have systems for very simple thing, like a sun sky system, which allows you to modify time of day and those kind of things. Do you see this approach to system we could extend just like you described for procedures, and we could declare parameters and let ... Or a drivable car? It's a typical entity in the world of a simulation and the world of games. A drivable car or a sun sky system. Do you think we could together work towards creating standards for that?

Guido Quaroni:

Yeah. It's tricky. When something is not fully understood, it's really difficult to get together and standardize it. That's why I said, "Okay, geometry, I think we're there. Shading network. Eh, maybe getting there. Rigging is getting a little painful. And now you're going into a full simulated system." So I think it would be a while before there is clearly the 90% of the definition. What is it? How do you define forces, collision objects and all those things to really set the condition for a simulated system in a standardized way. Again, right now we can barely say, "Okay, gravity and the mesh, that's the collision and a few other things." So maybe that will be at some point, but I wouldn't be surprised that initially there's going to be this race of figuring that out. And each one is going to say, "I'm writing it to make a ... " How many time you heard an engineer? "Well, I'm doing this, not for this, but I want to really make it that everybody could use it." So it's always that desire.

Guido Quaroni:

I can imagine this to be still a few years away, because I'm not clearly see a simplification or an understanding of it to the level of like, "Oh, I think we can sit down at the table and probably have a parameter space that is understood." I think is still too ... Again, rigging to me is probably the next thing to look after shading. And even that one is again, fully declarative in a way. You just define relationship and things like that. So that one, I, I don't know. Maybe who gets out there first and is faster and better, maybe gets an edge. But It would be, I couldn't imagine at a round table to try to come to an agreement

Patrick Cozzi:

And Guido, you're coming up on a common theme that we see with the podcast, which is how fast do we standardize. Right? And I think it's possible to be both too soon or too late. So I think you're pretty pragmatic. And to me, a few years feels totally reasonable to do a simulated car.

Guido Quaroni:

Yeah. Too soon, burns out. Maybe even a great idea gets burned out too quickly. And then obviously that if a company or people, a company or people feel like constrained by a standard when they still are discovering things. I think it's actually not going to hold it. And you know what, and people are not going to invest on it because "I don't know if it looks like the right one, I'll wait." Kind of thing. Right? So, yeah. I don't see people giving confidence on the, "Oh, we fully mastered it. Here it is." "Hmm. We feel confident in it." I think it's a ... And also kind, that's the exciting part. People are hacking around. That's great to see what they come up with.

Patrick Cozzi:

Yeah. Well, it's funny. When we had Vlad on the podcast, he talked about experimenting in the open as part of the key things to drive that innovation. Very well said.

Marc Petit:

Yeah, no. That's probably the right way to go about it, is to put a lot of ideas out there in public, get people, build up on top of that, and refine models from there.

Guido Quaroni:

Yeah, that's another thing that I try to keep thinking about it. What's the future look like? Especially when I look at Blender. Not because it's specific to it, but just the general thing about an open source system that by design is going to stay there, and there is a definitely that kind of a development approach versus the companies that are more like looking into keeping the software in-house. And how when things have not fully understood, what's the best approach there. Right? Internal "Should we just do it in the open?" Now at the same time, obviously that's what com companies are looking for, have that spark, and we have something and come out, and nobody even thought about it. Right now, I'm pretty excited to see what's coming out.

Guido Quaroni:

And I'm not trying to do a plug. I know it's going to look like a plug for Adobe, but I'm really excited what I'm seeing with Modeler. This is a new SDF modeling system that we are just announced, and it's in private beta. And it's very interesting to see. I can't wait to see what people can do in a different paradigm. Traditionally, I thought SDF was really good for organic sculpting. And I'm seeing so much hard surface going on with this tool that I'm like, "Wow, that's quite interesting." So be interested to see what that ... So that's an area where you want to do something maybe in a small group of people together, and then you open up and see what's the reception.

Marc Petit:

It was a very impressive demo. And I remember when I was working with the Foundry, there was a first attempt and doing 3D with technology and it was promising. They never cracked it, but looks like you guys may have cracked it.

Guido Quaroni:

Yeah. Again, I was a little bit unsure. Now I see that, especially when it goes into ... That's another part, right? Until you don't put your software in the hand of the artist, you can just fantastically all the beautiful things you can do with it. And this is a clear example where it wasn't exactly clear, and now when we're seeing the work. People are feeling like, "Hmm, okay. This it's a great team. It's like eight, nine engineers, then they're making something really ... I'm super excited for that."

Patrick Cozzi:

So Guido, if you took out your crystal ball, maybe five or 10 years, and given your excitement with the SDF, the signed distance fields, how is content created, right? Are we scanning the world with LIDAR and photos? Or is it all procedural? Is it all done with artists tools? Tell us.

Guido Quaroni:

Well, I hope it's a mix, right? You know, I think there is going to be a big chunk is going to be from real world. The question will be more about how do we build tools that can take even real world things and make it unique. Because to be honest, making movies and content from scanned thing is fine for Amazon selling goods. But for creativity, I don't know if it's ... What's the point of ... I've seen people trying to make digital movies that they can be shot in a real live action. And what's the point of that, right? So the advances of capture, and all the advances also in the AI space when it comes into animation and everything, I'm hoping that gives you that ... We used to scan pieces of cloth with this regular scanner, turn it into a texture and map it onto a surface. But the idea would be again, how do you merge the two? And also have having tools that can deal with that data? Because it's a different kind of data. It's unorganized.

Guido Quaroni:

And so suddenly you're like, "Okay, now I got this big mesh with things and everything. How do I turn into ... So I'd be interested to see tools that really based on that, versus ... We went from very, very structured nerves to poly mesh and subjects. And now even freaking dance meshes. And how do you deal with that? So, that's kind of my hope, that it's still a mix of the two, procedural and novel ways to alter and then scan. I think that tools to make you go faster, you save a bunch of time. But it's dependent on the purpose. For certain purpose as it is, for creating new things, a new world. Again, the metaverse that looks like real world, just with a nice illumination model or whatever? It's not that fun. It has to look really different to what I got in there. So that's why some will be captured, but some will be built hopefully.

Marc Petit:

So Guido, isn't that impressive? I mean, 25 years ago when you joined Pixar, if somebody had told you that that photo reality would come for free and you can spend all your money on art, directing the content. It would've been a dream, right?

Guido Quaroni:

Yeah. And either the real time. Years ago, when we thought, "Well, at some point you're going to create content, live ... I mean, he made the content live in the moment. I was like, "Really?" They're having a session with the director, where actually the decisions are made right there as you scrap timeline and you do changes. I always thought, "Okay, that maybe, who knows." But now I think there's so many production that are switching into doing it in real time. So now it's a matter of the quality. And the quality ... The thing about quality complexity has always also baffled me. But now maybe we're cracked and now we're not always behind.

Guido Quaroni:

We used to be always behind the curve. Complexity was making the timing not acceptable. But now, especially the latest I've seen with UE5. Hey, we're getting there. It's really big dance. Also because beyond a polygon per pixel, you're done. So there's going to be a whole ... I think I was thinking about another big challenge, right? Especially in the metaverse. How do we make sure that you're going to see so many pixel with your eyes, whatever they are. How do you make sure that, especially because you have to define it in data, it's not like real world? How do we make sure that we don't put more than a few polygons in each visible pixel, whatever that means. Because that's going to be one of the biggest challenge, for managing complexity.

Marc Petit:

Guido, you've been very successful. And one of the things is we like to ask our guests here is some advice. Because the metaverse opens up a lot of opportunities in many, many industries, but it's kind of this moment where many industries are converging to the same set of technologies. So what advice would you give to people that wants to get into and surf that wave of the metaverse personally?

Guido Quaroni:

Well, the first thing is to ... The feeling that the metaverse, whatever manifestation it is, it's actually something that you want to live in it, or deal with it, and experience it yourself. The excitement ... I worry a bit that this thing is so inflated. And the idea that the Ready Player One kind of a world, I'm not sure I'm a big believer. And somebody is, maybe that's the case. But the first advice is really to get into this world is going to be super fast, evolving, changing. And you may develop stuff, and then you got to throw it away, it's really starting from the passion. They want to be really building it for something better. For a beautiful future thing. The way for me joining Pixar was about, "I want to be part of that movie making process, whatever that takes." And not more about, "Oh, now it's the cool thing. Oh, I'm just working on the ... Do some metaverse."

Guido Quaroni:

Sometimes people get to into the trend and go there, and study should be really driven. Do we believe there is something there? For example, I believe more in AR than VR. Maybe I'm old, but I'm hoping of augmented the world where we have ... The idea of having avatar, I'm not a big fan of it. I would love to be able to see people. If I need to see or talk to a person, I want to talk to that person. Even via video, but that person putting something there. So I'm looking more for that kind of a world that is a augmented and gives you more, but it doesn't take away everything else that is so beautiful about the real world. So anyway, the passion, I will say that ... I'm sorry. Long, long winded answer. Being passionate and wanted to be part of it to make it great. Make it good, not just as a source of cool stuff or potentially money, or stuff like that.

Marc Petit:

Or following the hype cycle, which is absolutely intense right now on the metaverse.

Guido Quaroni:

Yeah. And maybe we have to think about it is, remember VR?VR got this big moment, and then it went dormant, and then it comes back. Now we're in everybody ... I've been seeing people talking about the metaverse I don't think any having idea what that means. But they have to say, "Oh yeah, because we'll do some metaverse [inaudible 00:38:00]." And so now I'm trying to think about, is this the thing and then it's going to go down dormant because we're not ready from devices? Oh, that's another thing interesting. When is the data that devices will consume 3D data? Will be also very interesting by design? Like the data that you send again is a 3D definition of the data, not images or how to lay them down and all that.

Guido Quaroni:

That may be another trigger moment, because right now devices that deal with 3D data, not quite. Even the VR thing, they get the video to be powerful enough. Right? So that would be an interesting moment, where on the wires ... That's what running on the wires.

Marc Petit:

So when do you think this first device is going to be?

Guido Quaroni:

Well again, the devices right now are mostly like there is some computer, but you get the images there. My thinking is the day that those devices actually they are onboarding 3D data live. And so then you got your location, your orientation, they have a 3D data there. And now it feels like crazy. Like, "You know what, I'm going to put all that stuff." But I feel like if you want to really resolve it to the maximum fidelity, you may want to go with that, with the 3D. So then you can those LOD, other stuff. But the resolution is happening on the device right now. Again is not quite. You're putting a phone on your face right now, pretty much. Right? It's putting a screen kind of thing ...

Guido Quaroni:

So maybe that's a technological, the amount of bandwidth. Images are fixed amount of bandwidth. I mean, given a resolution 3D data, it's unbound. And so maybe it will never be possible to do something like that. But that would be a moment, right? The moment, that's the way it works, that's why I was going back to that metaverse browser. By design, the link is actually a 3D description and the browser tried to figure out what to do based on a viewpoint. And not based on that's the page.

Marc Petit:

Don't you think this could happen on the cloud, and just being streaming the [inaudible 00:40:12] output?

Guido Quaroni:

Yeah. Maybe that's the reason because it's so complex, the only option is, yeah you send your location in terms of 3D data, but it's very minimal and you can get a fixed bandwidth that it will be images. Right? So that may be the ... But conceptually, the whole pipe is built on 3D.

Marc Petit:

So Guido, we covered a lot of ground today. Is there anything we didn't cover that you were hoping?

Guido Quaroni:

No, I think it was a great conversation. And again, I'm watching the ... I'm trying to figure out also how the different players, they're going to try to tackle this whole thing. Because you seen again. Is it going to be again, who conquest first? Is there actually space for everyone? And so making the assumption that there is room for everyone, so let's not worry about it. And then how do we get better and faster in that place? Where again, people can experience many environment and worlds made by different companies without one try to build a closed garden to try to fend off the other. Again, it happened in the past, when you tried to get there, tried to force it. And so maybe that's exactly the point of this podcast. You don't forget that.

Marc Petit:

I think we're all shaking our head in violent agreement here.

Guido Quaroni:

Stating the obvious.

Marc Petit:

And that's why Patrick and I do this podcast. I think we have to have this conversation. I think the intent is everybody says "The metaverse is bigger than any one company. We have to work together. Our contribution is to try to find a path, and create not only the conversation but hopefully we get people to motivate people to do things about it. On this topic, is there any organization or person that you want to give a shout out to that from your perspective, it would be a particularly important in creating the open metaverse?

Guido Quaroni:

I think this is really about the players. There are not that many players right now, the big ones that really can steer the ship. Right? There are the hardware vendors and the software side. So I may count them maybe 10 or less at this point. There are Epic and Unity and Nvidia in the kind of some software side. Nvidia's a little bit of both. And then there is Microsoft. The usual name, Facebook and all of that. So I don't see right now ... And actually, I like the fact that there is a bunch of players out there try to ... So again it goes back to room for everyone.

Guido Quaroni:

It's always tricky when you have hardware and software together. And so that's why sometimes it gets complicated, because one tried to privilege the other. And so that is a little bit of a complicated marriage there. But hopefully people will realize that this will not help anyone if you try to force it. "Okay. This thing only works on my hardware." Because at the end, that's really what makes difference for the user, for the consumer. It's easy to switch a link and change whatever software is running, but not easy to have a different device every time. That's actually an impediment to adoption for something like this, especially it's becoming wearable. "No, no. Oh, I guess, sorry. I got to take my glasses off. I need to put the Google one. Oh, sorry. I'm going to take the Apple now, double lens." That would be insane. So by design, actually we should think about, "Okay, the user's not going to change their glasses."

Marc Petit:

So do you see anybody playing a key role in potentially a shootout you want to give to create that open environment?

Guido Quaroni:

I mean, I see Adobe, I really see the value for them to really enter the creation of it. And then I think in terms of the actual who are the most equipped that right now, I think Epic and Unity are definitely. Because they have the most scalable and performance software out there. And so those are definitely the ones.

Guido Quaroni:

Then there is Apple because they're going to come with the whole device ... Sorry I can't really pick one. But I will say for shaping it, I think starting from the software will be the best place to be. Not from the hardware. I think about it, because the software is when you define the behavior and maybe you standardize it. And then you try to work with that the other way around, I feel like it's going to really concentrate it to like, "Well, that's what my battery, that's when my RAM, whatever. That's what it does." So historically I've seen better things coming from the software side. So maybe that's a little bit of an angle there.

Marc Petit:

By the way, when we asked Michael Kass a similar question, the answer was basically you and Pixar for open sourcing USD for ...

Guido Quaroni:

I think what we shouldn't put too much emphasis on ... I think it's a mistake to think that USD solve ... I think USD will find this place as other things are building on it, and really be able to leverage on it. And so I actually feel like, yeah it's right there. It's just happening there. Now there's so much else to try to learn from it. That's what I'm saying right now with the engineer years at Adobe. Some of them, they don't know USD. I said "You know what, I'm not going to impose USD right now. I want you to learn it and see what you can build on it, and it will take time." So that thing will go on its own pace. Yes, absolutely. If in five year, USD is still in one company and closed in, we have a problem, but it is not just a USD becoming really open and everybody can do, or is more of a consortium is going to solve ... I don't know if [NVO 00:45:57] is going to change to put that back into the code base. I want to challenge Michael, are you open sourcing that? Because that's slightly different.

Marc Petit:

Yeah. Maybe we should have got both of you on the same podcast.

Guido Quaroni:

Yeah. Fantastic work. So I'm super excited for their ... I mean, I was really pleased for their investment they made into the technology.

Marc Petit:

But I mean, on a very personal note Guido, I had the privilege through Autodesk to see how you drove USD. It took 10 years, right? It was 10 years ago and-

Guido Quaroni:

Yeah, 2013. It was the first time I remember. Almost 10 years, but I remember saying, "Hey, what you guys think?" Yeah.

Marc Petit:

Yeah, no. I remember that. But I think whatever happens, I have a prediction is, this work and the body of work that you led with your team at Pixar will be acknowledged as one of the most significant contribution to create this open world. This is my personal prediction Guido. I hear what you say about we're not done, but it's such a strong foundation, and I'm looking first to the day when we recognize you as the originator of that work.

Guido Quaroni:

Yeah. And I think it's more about the praise really goes to the engineers also. It was really great to see them to take also the due diligence and the rigor. And production was not happy, because they were going too slow when they was developing it. Right? They took it in a way to say, "If we have to open it up, it has to be really rock solid. It is almost like when Pixar makes a movie has to be a good movie." And so the credits goes to them, because I was lucky, actually to see them really wanted to make it such a good ... So it's up to us now to continue to support and improve it even from being an outsider now. And that's why I'm excited with Adobe to see there is definitely support and excitement to help the cause.

Marc Petit:

Yeah, no. This is great. Well, Patrick?

Patrick Cozzi:

Indeed it was, Guido. Thank you so much for sharing. Especially all the stories. Really exciting.

Guido Quaroni:

Oh thank you. It was fun.

Marc Petit:

Yeah. It was fascinating. So I want to thank everybody, our audience. We have more and more people every week listening to the podcast. And I think it's not because of Patrick or I, it's because people like you Guido, bringing their point of view to the podcast. So I have to do what every host has to do about remind people about subscribing, rating, raving, and renting, and reviewing and talk about the podcast. Let us know who we want to see there and what we need to do differently. So Guido, thank you so much for your time. It was a great pleasure. And Patrick, we'll talk soon. Thank you everybody.