#VRML V2.0 utf8
     Shape {    # no Material --> turns off lighting
       appearance Appearance {
         texture ImageTexture { url "test.mpeg" }
       }
       geometry Box {}
     } 

     Shape {
       texture ImageTexture { url "grass.png" }
       material Material { diffuseColor 0.1 0.8 0.2 }
       geometry ...
     }

Note that in order to use one-component textures and to turn lighting off, you can use an IndexedFaceSet with colorPerVertex FALSE (i.e., colors applied per face) and a Color node to tint the texture:

     Shape {
       texture ImageTexture { url "grass.png" }
       # no material specified --> turns off lighting calculations
       geometry IndexedFaceSet {
         coord Coordinate { point [ ... ] }
         coordIndex [ ... ]
         texcoord TextureCoordinate { point [ ... ] }
         colorPerVertex FALSE               # color per face
         color Color { color 0.1 0.8 0.2 }  # green-ish color
         colorIndex [ 0 0 0 ... ]  # use same Color value for faces
       }
     }

If you want to vary the color at each vertex or face (e.g., to add hue randomness), specify a list of different colors and apply to each vertex.

TIP: Beware of texture size limitations: It is critical to be aware of the specific texture mapping restrictions imposed by the rendering library of the each browser that you intend to use. For example, some rendering libraries require that all texture maps fit into a 128 x 128 resolution. Browsers will automatically filter all texture maps to this size, but produce blurry textures and waste valuable download time. Some rendering libraries require that the texture map's resolution be a power of two (e.g., 32, 64, 128). A conservative approach is to design your texture maps in the 128 x 128 or 256 x 256 resolution. Carefully read the release notes of the browsers that you intend to use before wasting your time on high-resolution textures.

Keep in mind that if the browser (i.e., the underlying rendering library) requires texture maps at a specific resolution (e.g., 128 x 128) and you provide a texture map at 64 x 128, you will have wasted half of the texture memory. Therefore, to maximize performance, use as much of the required texture resolution for the actual texture maps by combining smaller textures into a single texture map and use TextureCoordinates to map the individual objects to their appropriate subtextures. For example, imagine that you have one medium-size texture that represents a corporate sign, and smaller size textures that represent small signs or repeating textures in the scene, such as stone, grass, bricks, and so forth. You can combine many textures into a single texture map and allocate proportional amounts as you see fit (Figure 3-29). Note, however, that combining multiple textures into one image interacts badly with a rendering technique called mip-mapping. Mip-mapping relies on the creation of low-resolution versions of the texture image; these low-resolution versions are displayed when the texture is far away. If there are multiple texture maps in the original image, the automatically created low-resolution images will not be correct--colors from different maps will be averaged together. A similar problem can occur if you use JPEG compression, which works on blocks of pixels. Pixels from the different maps in the image may be compressed together, resulting in errors along the edges of the individual texture maps.

TIP: In general, if you are applying a nonrepeating texture to a polygon, the texture should have at least a one-pixel-wide constant-color border. That border pixel will be smeared across the polygon wherever the texture coordinates fall out of the 0 to 1 range.

Transparent textures in VRML act as "cookie cutters"--wherever the texture is fully transparent, you will be able to see through the object. An alternative is decal textures, with the underlying object material (or color) showing wherever the texture is fully transparent. Decal textures are not directly supported, but can be created using two different textures as follows: A mask must made from the full-color, four-component texture. The mask must be opaque wherever the full-color texture is transparent, and transparent wherever the full-color texture is opaque, with a constant intensity of 1.0. The full-color texture is applied to the geometry to draw the textured parts of the object. The mask is also applied to the geometry, effectively drawing the nontextured parts of the object. A two-component texture with a constant intensity of 1.0 is equivalent to a transparency-only texture map--the diffuse colors used for lighting are multiplied by 1.0, so the texture's intensity has no effect. This might be prototyped as follows:

     PROTO DecalShape [
       exposedField MFString texture [ ] 
       exposedField MFString mask [ ]
       exposedField SFNode geometry NULL 
       exposedField SFNode material NULL  ]
     {
       Group { children [
         Shape {
           appearance Appearance {
             texture ImageTexture { url IS texture }
             material IS material
           }
           geometry IS geometry
         }
         Shape {
           appearance Appearance {
             texture ImageTexture { url IS mask }
             material IS material
           }
           geometry IS geometry
         }
       ]}
     }

The cokie cutter texturing behavior was chosen because it is more common than decaling and because decaling can be done using two cookie cutter textures, while the opposite is not true.