Meshes used in real-time applications are usually composed of sub-meshes which contain vertices with different sets of attributes. This kind of mesh cannot be used directly in the current graphics pipeline architecture because meshes for interactive applications usually duplicate vertices to ensure that every vertex has a single set of attributes. This fact causes apparently contiguous surfaces to be split into pieces or patches, and so traditional simplification error metrics will fail in any attempt they make to simplify them. Here we present a method for this kind of mesh which is based on edge collapses and takes into account the information about attributes that contribute to obtain a more realistic appearance of the object, like norma...