The Case Study of Metal Gear Solid 4

[MGS.]

C11. Modeling Using Calibration
Bonneville T100 is a motorcycle made by the illustrious British manufacturer, Triumph. It also appears in the game. To create the data for this motorcycle, the team performed modeling using calibration (image-based modeling).

The team attached measurement markers on a motorcycle that they borrowed from Triumph. They took 16 photographs at 45-degree intervals in a concentric pattern. At the same time, the sound team recorded live-action engine, switch and starter motor sounds.

They used the image modeler to import the locator and camera information from the created data into XSI. For the imported camera information, they rotoscoped the same image as was used in the image modeler. Whether or not each camera position fully matched the model could also be checked from the image.

After setting the rotoscope, the team performed the modeling work. Because the camera position for each image was accurate, they were able to identify accurate positions. This enabled them to produce high quality models in a short period of time.

To express detail, they also modeled high resolution models for normal maps.

They used Ultimapper to create various textures from baked images. They then finished the data by performing settings for the real-time shader.

In this way, the team modeled the Bonneville in the game based on the calibration data. They modeled most of the other machines and items based on setting documents, using the XSI functions and working step-by-step. But also, because the gun textures vary according to reflections, they designed their own real-time shader especially for guns,to make it easy to perform specular gradation control.

Mr. Umatate had the following to say about modeling with XSI. "I think that XSI's modeling function is highly compatible with the production process for normal maps, which are essential for next-generation consoles. Often we created a low resolution model and then used the subdivision surface to create a high resolution model. In this process, modeling with XSI was intuitive and efficient.

Thanks to the XSI's non-destructive environment for data, we could perform trial and error operations to improve quality until we were satisfied with the results."