There are some interesting bits of information coming out. I have some here from the Content and Mesh UG.
This is Google’s famous tool for building 3D models. It has a low learning curve and quite a bit of ability. Using it came up in the Content and Mesh Meeting Monday.
It seems people are building things using Sketchup and bringing them into Second Life. All works well until they UNLINK them. A 100 prim Land Impact can jump to 500 prims. The reason is Sketchup makes a model as lots of small parts. When one unlinks the parts all the less than 1 prim costs that were being added together while linked, have nothing to add up with and thus round up to 1. The round up kills the prim savings.
Nyx Linden said, “…Sketchup exports to Collada in a ton of tiny pieces, so if each piece gets counted as a prim on unlink, it will cause that. We recommend that if someone wants to use Sketchup that they learn how to pass it through another package to combine/clean up the file before importing. Specifically, combining the pieces into sensibly-sized bits. This will also help reduce load on our end. Even when the parts are linked together, there’s a bit of tracking that has to be done for each piece.”
So, if you do decide to use Sketchup for your modeling, just know it is not going to create the most efficient models.
If you want to start with something simpler than Blender, try Wings3D. Also know that Blender 2.6 is way easier to learn than Blender 2.4 and previous versions.
Mesh & Prims
Mixing prims and mesh can save you Land Impact costs. But one does have to understand how usage can reduce or increase Land Impact cost. Using cubes and cylinders can create a savings. More complex shapes cost more with a torus being the most costly.
There is some desire to have the controls to Havok physics shapes exposed in the viewer. This would allow one to omit a physics shape and simply select the shape for the Havok engine to use. Nyx pointed out several of the problems. Mostly that it is not a small or simple project. The complication comes in collecting the information on which shape and how to scale and rotate it to fit with one’s model. For now that is most efficiently done by using a prim in the linkset and setting the mesh’s physics to None.
If one distorts a prim, the prim cost and physics cost can skyrocket. A twist or rotation are expensive distortions.
Also, the idea of having the Land Impact Cost provide a better adjustment for the cost of prims. When prims are CUT the poly count goes down, which reduces the load on the server and viewer. So, a cube that costs 0.5 prim in a mesh linkset could cost 0.1 for a 90% CUT cube. To ever get that Nyx says we will have to write a feature request. Someone will need to write a good user case and description for the feature request.
Davido Chrome has dug out the Havok instructions for best performance and placed them in the Knowledgebase. See: Best practices for creating efficient collision volumes.
Helena Lycia brought up the problem of determining what the Land Impact Cost is going to be when using prims as the physics shape. One has to upload and rez the thing to see what the actual cost is. A feature is needed to calc that value before upload. Nyx agrees but is not sure when the Lab will have the time to add it. But, once the feature added to the JIRA we can start adding WATCH’s.
Uploading complex models is problematic. It seems the upload gets confused as to which components are associated with which physics model… or part of the model… or something like that.
The uploader is sensitive to the order of things in the Collada file. Blender and the other modeling programs don’t have a way to control the order of the exported Collada. There is hope that some programmers interested in SL will fix the problem.
Cloudy, Blurry Avatars
This is a project that Nyx and Vir Linden are working with or at least keeping up with. The news is they have found several causes for the problem. I get they doubt they have found the main or all of the causes. But, there are some fixes now in QA for some of the minor causes. But, there is nothing solid to say what the final fix is going to be or when it will arrive.