Search form

The Digital Eye: Dissecting A Previs Roundtable -- Part 2

In this month's edition of "The Digital Eye," Bill Desowitz concludes his report on a recent Previs Roundtable discussion he moderated with a focus on videogames.

EAs Anupam Das explained that while designers prototype the story and figure out new technology, artists define new technology for rendering features. All Medal of Honor: Airborne images © Electronic Arts. 

Last month, I provided an overview of an unprecedented previs roundtable that I recently moderated at The Art Directors Guild. Given that GDC has now concluded, its appropriate to end with the discussion of videogames by Anupam Das, lead artist at Electronic Arts. Das, who migrated from film as a TD (National Treasure and Phantom of the Opera), and is currently working on Steven Spielbergs eagerly anticipated LMNO game, gave a presentation of his previous project, Medal of Honor: Airborne.

Das began with the proposition that that gaming is a much more open world environment than film. How do you plan that? How do you know how much time and technology its going to take? How do you know what the feelings going to be, both in terms of story and what the players going to do to make the game fun?

And while the designers are prototyping the story and figuring out what new technology is required in terms of AI or navigation or physics or the construction technique so it will work on a real computer, the artists are in charge of defining new technology in terms of rendering features, what kind of world is being created and how to prototype the scale and mass of the world.

A lot of artists in games are architects in the sense that we are designing a world, Das continued. Were designing a space. And much like architecture, we need to define our contact before we go deeper into the details. And then we prove that our environment, our level, can run at a target fringe FPS, a target frame rate. Generally, thats about 30-frames-per- second. Some games actually shoot to like 60-frames-per-second, like Halo. The challenge is to get a high degree of fidelity visually, a high degree of interesting game play, all working on your target platform... So, ultimately, we want to make sure that the game is fun, and we want to figure out how to do that with a minimum amount of work before we go into full production and enter previs.

But rather than using previs as a reference in games, Das uses it to morph the actual assets and technology into the final product. And for Medal of Honor, they created a very nonlinear world where you could go anywhere.

It was based on the 32nd Airborne Division, so you could actually fly in an airplane, jump out and drop anywhere in the world and join a flight, Das explained. So we wanted to also have a living world, so instead of having guards everywhere, we had artificially intelligent characters that could understand where the tensions were and where battles were fought. They could see things. They could hear things. They could communicate with each other, so it was a real living world, and thats one of our targets.

Being a World War II game, Das wanted to retain historical accuracy in terms of the storyline and the environments that were created.

And being a World War II game, we wanted to retain historical accuracy in terms of the storyline and the environments that we were creating. So first thing we needed to define, and this is kind of the first step to pre-visualizing our environments and our game, is to know what the story is or know what the context is. So we looked at a lot of reference of the era during this war, the kind of architecture, the kind of people that live there.

We even hired a historian to help us define some of these things. We looked at more graphic, iconic sort of things, like, for example, silhouettes, atmospheric effects, a lot of the kinds of traditional stuff that you think and film, but it really needs to be culled out. I still think the challenge in games is that we have a responsibility to push it more toward the cinematic goal, that its not just enough to have a compelling story... you need a believable world that youre sucked into. And currently I cant think of one game thats done both in equal parts, and that is a challenge. So we call these specific things out and dive into it right away from the perspective of the space and in terms of the color palette and the shape language.

Das then proceeded to go into detail about the design and how it dictated the technology, which, in turn, dictated the environments.

We rely on production design and try to visualize some of those ideas [about fighting on rooftops and creating different styles of rooftops in Italy], just in terms of basic concepts and really quick concepts. And then we make a 2D map, which outlines the game flow, and its like a top down city plan view. And then from there, we block out the level using SketchUp, and then once we went to using Unreal Engine, we were able to actually block the stuff out within Unreal, and we took SketchUp out of the picture.

But SketchUp really is a pretty powerful tool for just prototype environments. So then, once we went to Unreal to prototype our environments, we could make iterations really quick, and designers were able to tune the space so that AI navigation would work, that youd have control of sight lines, that youd control the flow of the game and basically you could put a couple of guys in there and play it.

And even though there are no textures or any final lighting, you have a sense of, OK, I think this levels going to be fun. Were able to prove out some of the things that we wanted to do way in the beginning during the pre-production phase. So then, on the art side, we pick out color palette, textures and then we pick a key area in the level, and the level is huge.

After bringing the image to a high level of visual fidelity with lighting, textures and shaders, Das said that reaching the target frame rate is the next decision. How much CPU is being taken up by the world they are rendering?

And after bringing it to a high level of visual fidelity with lighting, textures and shaders, Das said that reaching the target frame rate is the next decision. How much CPU is being taken up by the world they are rendering? How much of the GPU are they taking up? This is where something called PerfHUD, which tests the space for target memory requirements, comes in handy.

So this is why the previs is so important, because before you go wide and just build a really cool, great, filmic environment, its got to work at 30- frames-per-second. Otherwise, its going to be no better than a PowerPoint slide, right? But if its not going to run at frame rate, you have to remove stuff, and thats the worst situation. The best situation is you plan ahead, and you dont have to remove that and affect your visuals.

Is it too expensive on the CPU? Generally, theres a tradeoff: you can do a lot of stuff that fits in memory, but then it has a bigger impact on the CPU and vice-versa. This is what we call DPC coloration, so the more objects you draw in Unreal, the slower it is, basically. And every single unique color is an individual object. So at any given time, we have over 30,000 unique objects being drawn on the frame.

Thats pretty hefty, and its amazing that todays hardware can do that in realtime with shaders and lighting, dynamic lighting with dynamic soft shadows and ambient occlusion. So its a pretty big challenge.

In a follow-up exchange about convergence and the challenge of pushing technical and emotional boundaries on LMNO, Das offered, On LMNO, its much more story-driven, and so the way were tackling the previs of that is... just prototyping things very quickly. So, for example, we might have a small storyline, like a paragraph that explains something that happened. When you enter a room, you meet someone, and some story unfolds. What well do is rough out some really basic animation, some really basic geometry, some audio and well just build it in Unreal.

Das finds the challenge is integrating the story with game play. A really compelling story is not fun if the players not doing anything. 

Theres a ton of game engines out there. Theres a ton of talk about various types of rendering technologies out there. But Unreal allows you to iterate rapidly. I literally can build an entire level and prototype it in about two days, which is pretty ridiculous, because two years ago, before Unreal went wide, that process would take about three to four months in my own experience.

"I think the problem is integrating the story with game play. You know, you could have a really compelling story, but if the players not doing anything, its not fun for the player. Thats an ongoing challenge.

Story is important. Moving the fiction along as youre playing the game play is important. [But] we know what film brings in terms of emotional context and experience, and thats equally important. Its about bringing those two disciplines together. Thats the hard thing. The solutions not so apparent right now.

The thing that we need to do is not be afraid of trying it, and its going to take a lot of money, but its like in film: people are trying to render realistic humans. Well, I dont think that anyones really nailed it, but it doesnt mean that we shouldnt try... And so I think LMNO is a step in the right direction of trying to prove that you can create emotional context in the realm of a videogame, and make it fun for all ages.

Bill Desowitz is editor of VFXWorld.

Bill Desowitz's picture

Bill Desowitz, former editor of VFXWorld, is currently the Crafts Editor of IndieWire.