Good conversation on the pros and cons of The Mesh Generator from Naonao Watanabe, the tool for converting prim-based builds into mesh in Second Life, from two longtime builders. First some advantages (and tips) from Han Held:
I've been using the Mesh Generator tool for roughly three or four years now. It's been a huge lifesaver for me, and there have been a number of times when I've managed to get significant LI savings by using it... Even though I am playing with Blender and related tools (like Avastar) the scales haven't quite fallen from my eyes yet. Mesh is still a confusing jumble for me.
One thing I'm doing with generated meshes which seems to help some is to import them into Blender and then clean them up there. Using the "mesh" > "clean up" submenu in Edit Mode and then hitting the object with "degenerate dissolve" and "decimate geometry" and re-exporting it.
I only usually end up doing that for OpenSim, which has become more finicky about meshes since the release of 0.9. But occasionally when the LI is through the roof in SL that procedure has helped to bring it down to earth.
That said, there are still a number of times (like 1/4th of the time) where I find out that the LI of the Mesh Generator-created mesh object isn't that much less than just simply prims ...in that case, I just stick to prims.
Some caveats from Chic Aeon, referring to the tutorial video featured in the original NWN post:
Nice clear video BUT:
BUT? LOL. That door is 598 triangles. For any kind of optimal mesh it needs to be taken into a 3D modeling program and have MANY vertices removed.
A prim has 48 triangles (originally for more "lifelike light reflection"). A MESH CUBE has 14. So this final door (as shown in the uploader) has WAY too many verts. The only advantage of this method would be to lower land impact -- and it certainly can do that.
But for many people that want to build with prims, just turning objects to convex hull will cut land impact dramatically (approximately half). In both cases you have the issue with physics not being correct on some objects -- as in the door frame in the video.
We see lots of questions about this on the SL forums. One BIG issue is the physics model, another is texture rotation. The list goes on. Several techie devs have said that mesh made in this way is LESS efficient than prims as simple prim shapes are built into the viewer and easily recognizable.
I wasn't exactly sure how much more efficient this door would be made directly in a 3D program and so I made it. My version is very similar to the door in the video. (I would have actually made it a bit more "elegant" if it was a real door I was going to use). One door knob not two and no hinge -- which is necessary for most door scripts to work on a mesh door.
My Blender door came in at 168 tris.
You can see it with the numbers here:
So I would like to encourage folks that DO want to make efficient mesh, to put in the effort and learn to make it within a 3D program. The list of reasons is oh so much longer than my short post here :D.
So overall, seems like the Mesh Generator is very valuable as a way of learning how to create mesh in SL and works well in many instances... but performance-wise, not so much in other cases.
Great article. It got me searching for LSL scripts that I might use to create my own Prim to Mesh tools; I'm always looking to build it myself before buying a tool like that-LOL. In my search, I discovered that Firestorm viewer has a completely free Collada exporter built into their Firestorm viewer:
https://wiki.phoenixviewer.com/collada_export
It will export a linked set of Prims to a Collada mesh saved to your hard drive.
Since you already have The Mesh Generator, you might want to compare it's output to the Firestorm output and see if one is better than the other.
Thanks for the excellent article!
~Sean
Posted by: Sean | Saturday, August 11, 2018 at 08:34 PM