Every polygon added to a real-time mesh takes a certain amount of time to render.
Even if it is less than a 1000th of a second, that time adds up and diminishes the possible frame rate during game play.
Texture mappings are drawn much faster than the polygons needed to create the details that could be "painted" into the Texture map.
To be effective, an object's material must simply be able to fit into RAM.
Therefore, any detail that can be effectively simulated by adding it to the material should be mapped, not modeled-as should any detail that is too mesh intensive to be effective (such as the "cookie cutter" trees discussed earlier).
A great example of when to map instead of model is the muscle tone in a character.
Nice, rounded muscle structure is far too polygon intensive to accomplish in real time.
When muscles are added to a texture map, however, a similar effect can be achieved through careful use of simulated highlights and shadows, with almost no cost in frame rate.
A helpful process when you create real-time models is to create a fully detailed, high-face-count model first.
Then construct the low-resolution model over the high-res model, using the latter as a template.
You can then take individual orthographic renderings from the high-count model, tweak them in a paint program, and then use them as texture maps for the lowres model.
This process is discussed in more detail in the "Dealing with Texture Limitations" section later in this chapter.
Another related process is to load two images (preferably scanned pictures) of the object you're creating, seen from the side and the front, as a texture map to be placed on a two-quad "tree" in your 3D Application.
This tree can then be displayed as a shaded template to be built over.
This technique can be very handy as a reference for building real-time characters.
Even if it is less than a 1000th of a second, that time adds up and diminishes the possible frame rate during game play.
Texture mappings are drawn much faster than the polygons needed to create the details that could be "painted" into the Texture map.
To be effective, an object's material must simply be able to fit into RAM.
Therefore, any detail that can be effectively simulated by adding it to the material should be mapped, not modeled-as should any detail that is too mesh intensive to be effective (such as the "cookie cutter" trees discussed earlier).
A great example of when to map instead of model is the muscle tone in a character.
Nice, rounded muscle structure is far too polygon intensive to accomplish in real time.
When muscles are added to a texture map, however, a similar effect can be achieved through careful use of simulated highlights and shadows, with almost no cost in frame rate.
A helpful process when you create real-time models is to create a fully detailed, high-face-count model first.
Then construct the low-resolution model over the high-res model, using the latter as a template.
You can then take individual orthographic renderings from the high-count model, tweak them in a paint program, and then use them as texture maps for the lowres model.
This process is discussed in more detail in the "Dealing with Texture Limitations" section later in this chapter.
Another related process is to load two images (preferably scanned pictures) of the object you're creating, seen from the side and the front, as a texture map to be placed on a two-quad "tree" in your 3D Application.
This tree can then be displayed as a shaded template to be built over.
This technique can be very handy as a reference for building real-time characters.