The Case Study of Metal Gear Solid 4

[MGS.]

C12. Background Modeling Process
At the beginning of the project, staff from the background group went to various locations both inside and outside Japan to shoot materials. In March 2006 they went to Morocco, in April they went to Peru and in June they went to Prague in the Czech Republic. In Japan, they shot photographs of steelworks and of giant flood containment silos under Tokyo. With these shots, the team wanted to capture the atmosphere and details of their surroundings, but they also wanted to use the photographs as materials and for creating texture. As such, they took over 35,000 photographs.

The background modeling process was as follows. First, the script group in charge of the map design created the rough model data. The data were given to Mineshi Kimura and Yutaka Negishi, who developed realistic data by correcting any parts that were structurally or pictorially inappropriate.
They returned this concept model to the script group, who checked the game balance.
Then, the team performed the actual modeling editing and material setting, and set the collision information. This brought the data close to completion.

Background data in the concept map stage that was submitted from the script group.


Background data with initial stage corrections added by the background group.


The final modeling stage.
Texture setting and detailed modeling is performed in parallel.


Collision model data.
These are quite rough data that are used as reference for players, enemies, animals or vehicles.


The final image on the PS3.

The team did not use LOD for the background data. They made partial use of LOD only for large volumes of trees or grass. 150,000 to 200,000 polygons were imported into the PS3 in each load. If the data exceeded 200,000, they were divided into separate loads.

Even when a low number of objects was expected in a scene, sometimes there were more than 5,000 objects, and sometimes the number approached 10,000. The team used XSI's NetView to manage such large numbers of objects. First, they created a management page for a small object list. Small objects registered on this page, such as bottles, frames and potted plants, could be easily positioned just by dragging and dropping them to the XSI background data.

A wide variety of attribute information was set for the objects that appear in the game. For example, objects for when guns are shot, sound effects when something is hit, physical properties, and OctoCamo information (changes that camouflage Snake's suit as he creeps alongside a wall). These attributes were easy to define on XSI by giving them names. The team added a Null inside the object node and then set the name, such as IRON_*** or WOOD_***. When these data were output to the game, the various reactions that were specified by the attribute information were shown.

Example of OctoCamo changes.

The team performed the advance baking of apex colors and light maps with XSI's render map. In general, they used Final Gathering to bake rough shading into the apex color. However, in the case of light maps, after significantly compressing the texture volume, this technique is not required in areas where the shadow buffer is active. The shadow buffer can be used much more in the PS3 than in previous consoles, so the background shadows and character shadows can co-exist in a non-destructive way. For this reason, in most cases real-time processing is used for shadows.

But in the shade, the normal map is less effective than in sunny areas. Another issue is that second and third order light expressions are also required, which the shadow buffer cannot keep up with. In response, the team set data known as LSC data to supplement the lighting texture. LSC data stores light positions and brightness information in an external file, which can then be used on the console. They performed the lighting settings by importing the background created with XSI into a lighting editor, an in-house tool that displays a preview that is almost the same as the image on the console. After adding colors and accents in a way similar to painting, they could perform a rough preview even while in the OpenGL condition. They could also perform quick and accurate previews on the console by compiling from the lighting editor.

In addition to lighting for these background models, the team also performed lighting for the characters. They set the point light sources and environmental lights to suit the atmosphere and colors of the background. After finishing this, the final job was to perform the settings for the OctoCamo. The game is notable for how the light intensity changes in different situations, such as for camouflage, in direct sun, in the shade and when light bulbs are broken.

Work screen in the lighting editor.

Comparison of lighting effects on a character.