#SL Mesh Deformer Status II Week 48

There is an issue for some with the Mesh Deformer and the time it needs to calculate the mesh for offsetting. Kitsune Shan has been working with the Deformer and voicing concern about the time needed for the Deformer to kick in.

Kitsune Shan

Kitsune is proposing that we do something like what is done with avatar baking. The viewer does the bake and sends a composite image to the servers for everyone else to use. Of course that is being changed to a server side process now, where a new service in servers at the Lab will do the bake and send out the composite image. Kitsune would like to see something similar to that for the Mesh Deformer.

White Rabbit points outthe difficulties with attempting to cache the deformation calculations. White Rabbit has done some testing and finds the CPU time for the Deformer calculation is consistently small. She speculates that the long calc times come from the thread being a low priority process.

White Rabbit

The chances that Deformer calc’s will be cached server side is, in my opinion, slim to none. We do not have any statistics on the type of hardware being used with Second Life™. We know there is new stuff, way old hardware, and everything in between. But, AFAIK, we have no information on how much of each.

We have been told that number of users staying with SL Viewer 1.23.x is only 2 or 3%. Third Party Viewers that do not use SSE2 CPU commands is small. Read the SSE2 thing as meaning viewers that support really old computers. Their user bases are so small they fit within the 2-3% too. So, I’ll speculate that the Lab is not going to put much effort into something as complex as caching deformation data.

Collision Bones

Using Collision Bones to help with mesh clothes fitting is an ongoing discussion. At Monday’s (11/26) Content and Mesh UG meeting the subject came up again. It seems some designers are looking at using the Collision Bones as a means of getting a better and adjustable fit.

We first saw the idea back in late June. See: #SL Mesh Deformer Alternative. The process works, sort of. Some people use it. It is Deformer independent, so it works now.

But, using collision bones has limits. For one there is no bone to handle breast size. Breasts cannot be sized with collision bones as they will be with the Mesh Deformer. Changing your shape will not work for breasts when using collision bones. Nor will clothes made using collision bones respond correctly to Enhanced Avatar Physics (bouncy boobies). These are issues the Mesh Deformer handles.

During the user group meeting Nyx Linden made it very clear that while the Lab is not prohibiting the use of collision bones for mesh deformation the Lab is not supporting that use of collision bones.  So, at some point if a change was needed they would not have any aversion to breaking that non-feature.

Alternate Weighting

Along with the Mesh Deformer is discussion of JIRA item: STORM-1800The vertex weights of the default character mesh could be better.

Darien Caldwell did some experimenting to see how the alternate mesh and weighting would work with the Mesh Deformer. This whole line of thinking is part of the Avatar 2.0 thread of thinking. STORM-1800 is sort of what I call a request for Avatar 1.5.

STORM-1800 is about making repairs to the avatar that would help but not break existing content.

Avatar Bend Problems

Darien describes the test results, “I was actually pleased to see it didn’t have any noticeable effects in that department [Mesh Deformer]. Most of her [Alison Alena] weighting changes seemed to have visual improvements to the stretching on faces when bending, but didn’t actually change the way the avatar bends, which I found a little odd in reference to some of her pictures, especially on the stomach. She claimed it helped make that area bend cleaner, but I didn’t see any change. It may be very animation-specific.”

The image shows the SL Avatar deforming as it breathes. You can see a part of the avatar folding under, the missing yellow parts of the line. The problem is from a combination of the avatar mesh layout and the weighting.

To see how the changes from STORM-1800 work, use the Dolphin 3 Viewer. As of version 3.3.18 (August 2012), the Dolphin Viewer has used the modified weighting.  This is not something you can enable or disable. It is apparently just built into the viewer or not.

These changes may actually be part of the SL Viewer… someday.

Summary

It looks like the Mesh Deformer may be hard on some older computers. Well, made mesh clothes will reduce the problem and poorly made clothes will aggravate the problem. All the indications are this calculation delay will be part of the Mesh Deformer when it arrives on grid.

For reasonably new computers and those using reasonably well made mesh clothes the delay will be measured in tens of seconds or more likely less.

Collision bones are considered a short term work around while we wait for the Mesh Deformer. When used for deforming clothes for better fit the Lab will not be a supporting the use. That means that at some point clothes made using collision bones may break. The keywords being ‘may break’.

Just for kicks look at: http://www.plurk.com/p/hq4s8m

2 thoughts on “#SL Mesh Deformer Status II Week 48

  1. Pingback: SL project news week 48/1: server and beta, viewer, maps and memory | Living in the Modem World

  2. Pingback: SL project news week 48/1: server and beta, viewer, maps and memory | modemworld

Leave a Reply

Your email address will not be published. Required fields are marked *