Reducing Poly Count
We will get into the reasons for why we need lower poly count models of our model later. For now the tools that come with Bsurface in Blender 2.6 make modeling much easier.
A handy tool for working down poly counts is DISSOLVE. In the image I have selected 4 faces and pressed ‘X’ and selected DISSOLVE to remove the faces. The action turns the four faces into a single ngon.
The action of Dissolve is dependent on what is selected. If an edge, then edges are removed. Of vertices are selected then they are removed.
You have considerable freedom in what you remove. It makes reducing poly count much easier than previously. The trick to using Dissolve for Second Life is to use Edge Select and Dissolve to avoid leaving behind ngons. I could do a whole tutorial on just this feature.
Along with Dissolve is Edge Collapse and Delete Edge Loop. Using these takes some learning to anticipate what they will do. So, experiment with them. Make a cube or iconosphere. Subdivide the cube 5 or 10 times so you’ll have edges to work with. The iconosphere is ok as is.
Many modelers use an image drawn by an artist or a picture of an object to guide their modeling. There are a couple of ways to use images. The original Blender feature is Background Images. You’ll find the feature in the Objects Tools (‘N’ in 3D window). Look for Background Images. You can move and scale the image from controls that reveal once you add an image.
There is a new feature named Image as Planes. It allows you to do more with the images. Watch the tutorial to see if it is something you want to use and how to use it. See: Blender 2.6 Image as Planes.
In blenderCGartist’s tutorial he uses the planes in an ‘L’ shape. For a character it is often easier to put them in an ‘X’ arrangement. Do what works for you.
How to Model
For this tutorial we are considering how you build your model rather than what to build., which is the ‘what to build’ decision. An artist’s first inclination is probably to think about how the final product will look, style, shape, and etc. For modeling we need to anticipate how the model will move when animated. We must make places for the model to compress and stretch.
So, the artist will often create the model to look as desired with few other considerations. After that model is made, it is re-modeled to provide a topology that will work with animation. This is why we see so much emphasis on retopologizing (like that’s a word) models.
We also need to redo the topology to create the lower levels of detail. In Second Life the levels of detail are referred to as LoD. Modeling in Second Life uses 4 levels of detail. The LoD has a huge effect on the Land Impact cost of mesh objects, which should tip one off that there is a high render cost for high poly models.
In Second Life the LoD’s are labeled High, Medium, Low, and Lowest. These levels kick-in at different distances. The HIGH LoD has the most polygons and the LOWEST LoD has the least.
The distances at which the LoD change depends on several factors including viewer settings. The default values are described in the Second Life Wiki: Mesh Streaming Cost. The size of the object and the camera distance from it decide which LoD level is used for the render.
For clothes I like to use the same model for HIGH and MEDIUM. For objects I try to reduce the MEDIUM LoD poly count by at least 25%. For the LOW I reduce the poly count as much as I can while keeping the silhouette. In the LOWEST LoD often two cubes with a picture of the object will suffice. It all depends on the size, shape, and distance one sees when looking at the object. Done well mesh cost is less than prim cost. If you are not finding that to be the case, start asking questions.