Darien Caldwell did a test of the newly released version of the Mesh Deformer. The test is shown in an 8 minute video posted on YouTube. You can discuss the test over on SLUniverse: SL Deformer Test, where I first saw the video.[youtube 1wnBB7c5NfM]
In the video Darien has the avatar wearing a copy of the clothing templates so we can see the avatar’s polygons. He used a copy of the avatar as a rigged mesh attachment with a black texture for clothing. This makes it really easy to see the places where the two meshes match up and overlay each other.
The places where the white and black flicker are the places where the avatar and attachment polygons are so close to being in the same place, the viewer is having a hard time deciding which is on top and visible.
His testing is pretty extensive. The video runs 8+ minutes. You can see how the Deformer works as Darien changes appearance sliders. In general it does pretty well. But, there are some problems. These are likely more a result of the nature of the SL Avatar than it is the Deformer. However, in the JIRA STORM-1716 some developers think there are problems with this version of the Deformer.
Once the attached mesh falls inside the avatar things tend to go wrong. At one point Karl had talked about doing some checking to see if the attached mesh was going inside the avatar. Of course if such a check was done, how would that affect custom avatars that do not conform to the base avatar and need to be smaller than it? So, that idea may have been rejected at some point.
I think the test shows that we are going to have problems when shapes reach extremes, at both ends of the scale: small and large.
At time mark 7:30 Darien tests the Avatar Physics… boobie bounce. It works.
Darien did remove weighting on the face and hands, according to his post at SLUniverse. That does stop the hands and face from deforming. So, the lack of deformation in those places is because of how the mesh was weighted not how the Deformer works.
Siddean Munro points out some problems;
My expectation is simply that if you overlay a mesh that has the same number of vertices in the exact same positions as the SL avatar mesh, when you move the sliders, the new mesh moves as the SL avatar skin does. In this version, when you slide the torso muscle slider up, the arms on the test mesh go down. When you slide the torso muscles down, the arms develop rather unattractive bulging wings. The lower half of the breast doesn’t deform, but the upper part of the torso does when you change breast size. The feet are all over the place – worse when you have a shoe shaper on. The fingers on the test mesh are pulled in strange directions instead of conforming to the shape underneath. The ear sliders deform the entire side of the head instead of the ears. The ears don’t actually change size when you make the ear slider larger or smaller, but the skull does. Darien showed the same errors in the movie he made of his experiment today. I’m kind of hoping we’ll get a response from Oz or Qarl on this 🙂
I’ll remind you that Darien removed some weighting in the head and hands. But, that should not affect the other issues.
I’m not sure that ear sliders, shoe shaper support, and hands will be that important. I could be wrong, but those are areas where clothes are not worn and have no need to fit that well. Shoe shaper support is going to be a matter of taking off the shaper.
Foot size is more of an annoyance. I would probably design shoes using an avatar base shape with the foot size set to zero. I think most females in SL have their foot size set to zero. Mine have been since I came to SL and I have only ever changed that setting for testing the Deformer. So, designing for that size seems reasonable to me.
Still it is a pretty interesting test. Thanks Darien.
In most 3D real time animations mesh models are not overlaid. The mesh body of the avatar is usually removed where covered by clothes. In SL we use Alpha Layers to hide the avatar mesh because we cannot remove it.
In trying to make blouses and skirts look realistic we are doing something seldom done in 3D virtual worlds. Most worlds have a very limited set of avatars and clothes. These are mostly made by professionals and added to the game.
In Second Life we are creating a much more generic system that must handle a wide variety of user created mesh objects and attachments. Most users are not professional 3D modelers. So, all the problems are going to show as we won’t have the skill to hide imperfections in the system.
Some third Party Viewers are using Avatar 1.5, which is my versioning where the current base avatar is Avatar 1.0. Or they are providing ways for users to add in Avatar 1.5, which has better weighting. I’m unclear whether some have better mesh topology.
The problem with the Avatar 1.5 solution is that only those that have added the upgrade see the difference. Technically, this breaks the Linden Viewer Policy of maintaining the Shared Experience. So, any TPV’s with the feature are likely going to be relegated to experimental versions.
However, users can add the feature and not be concerned about policy just as they would add custom Windlight settings. It is only something the individual user sees.
It is looking like the Deformer is pretty much complete now.
Testing remains. Some think the Lab should create the test clothes and get the project completed. Often these are the same people that complain about how poorly the Lab’s staff actually builds in-world and how little they know about SL and how it is used.
I think leaving the testing to the Lab is a rather scary idea, no offense intended. But, 3D modeling and fashion is not the Lindens’ forte. So, I am hoping some designers contribute new testing clothes for the new Deformer. Especially those designers that have run into problems. Clothes that exemplify the problems are the best clothes for testing.
PS: If you use Cool VL Viewer, Henri tells you to: For the Cool VL Viewer, you need to enable the experimental mesh deformer feature in “Advanced” -> “Rendering” -> “Enable Mesh Deformer”.