Misconception

There is a concept that made it into the JIRA that the base shape of the avatar is not important. The thought is that the Deformer just fits the mesh to the avatar’s shape. I suppose the idea is that the Deformer works mesh like it is spandex. It moves the mesh of the clothes to fit and follow the avatar shape.

The idea is wrong. We know because Karl tells us so and by a little thinking. How does one make a baggy shirt or a tight fitting blouse? Something has to tell the computer how to fit these different things to the avatar. I’ve drawn a picture to hopefully convey the idea of the Deformer calculation and some of the problems. I won’t swear it is correct, but it should be close enough.

Mesh Distortion Problem - Click to ENLARGE

In the image immediately above I’ve tried to show what happens. Consider the inner, brighter mesh the Base Shape. The outer, fainter mesh is the custom shape, one modified by us using the appearance sliders.

A designer makes a top, represented by the dark purple line. In the image we can see the distance from ‘B’ to the purple line. This is where the system decides how far off the skin the blouse should be. Next the deformer has to calculate how far it is from ‘B’ in the Default Shape to ‘A’, which is the same point on our custom shape. I think Karl can figure that out because both meshes are the same and the vertices are in the exact same order.

Next it adds the distance from ‘B’ to the dark purple line to create the position for the light purple line. But, if you look at the opening image, you can see something went wrong. The problem comes from the mesh making up the blouse not being identical to the avatar’s mesh. Karl has to guess which blouse vertex to apply the calc’d offset distance to using.

This is where we run into problems. It is easier for me to show what is happening using ‘A’ and ‘B.’ We can see in any given picture pretty much which point should be moved. We know ‘A’ corresponds to ‘B.’ But, in the blouse we don’t. All the computer has is a list of numbers. It has to use some process to figure it out.

Generally what works is finding the closest vertex. But, ‘B’ is not the closes vertex to ‘A’ after the shape change. Now ‘C’ is the closest vertex and it is a different distance away and shorter than what we need it to be to keep the blouse smooth as the designer intended. So, we get a glitch like we see in the opening image.

Karl added smoothing that looks at more than just the single closest vertex. That helped. But, there are still places where it is just not going to be adequate.

The greater the distance from ‘A’ to ‘B’ the worse the error is going to be. The current thinking is the smaller the distance is, the smaller the error will be. So, some of us think the Alternate Base Shapes is the solution. By having base shapes that more fit closely to the custom shape in use the offset distances the Deformer works with are kept smaller keeping errors smaller.

A Problem

One of Karl’s comments on Qarl’s blog is pointed out in the JIRA: “There’s been a proposal to allow for alternate base shapes – a male base, and possibly multiple sizes for both male and female. Sadly this work was not scheduled in the original bid, so it’s unclear how it will be funded.”

I have no idea how much work making the changes would be. As I think through how I would get a set of X-Y-Z points that represent a base shape into everyone’s viewer and tell the Deformer which to use it gets complicated. So, I can’t guesstimate the cost.

What I do know is Max is on the One Size side of the debate. So, we probably should not expect him to step up for a funding effort.

Funding is a problem.

Add in the Lindens

Karl pointed out in the MetaReality podcast and in the JIRA that the Lindens are thinking about what would be the best shape to base the Deformer on. In responding to Elie he said, “I do not believe there is anything inherently evil about having multiple base shapes. Internally, Linden lab is concerned that the default shape may be a poor choice for the base, for very similar reasons.” (Reference)

This means the Lindens are considering the issue too. There is some pressure to get the Deformer out the door. If the Lindens decide on a shape, we all may find it disagreeable. I expect the Lindens to look at the solution from a practical and system efficiency viewpoint. That may not get me what I want.

Plus most of us are gearing our design work predicated on the information that the Default Shape is what we should be modeling on. So, the longer the Lab puts off a decision the worse the pain will be.

If you are interested in the outcome, get involved.

Summary

It seems the One Size Deformer people think the Deformer will work OK and the texture and mesh distortions will be insignificant or can be compensated for by designers.

The Alternate Base Shapes side thinks the Deformer is limited and distortions will be significant for all but the most average shapes. So, for them the multiple base shapes using standard sizes is a good solution.

Around the edge of the argument is Avatar 2.0. I am convinced the Lindens are NOT going to go there. In the podcast Karl points out the community has never successfully forced anything on Linden Lab. I’m not sure that is completely accurate, but it is probably close enough for this debate. The Lindens have a lot of resistance to any idea of changing the avatar. So, I think Avatar 2.0 by the Lab is moot point.

Karl points out, again in the podcast, that Alternate Base Shapes possibly using other avatar meshes could give us a way to implement an Avatar like 1.8. It could work for elephant and dragon shapes too. It would certainly give designers way more freedom.

To me it looks like the Alternate Base Shapes extension to the Deformer is a great idea. I support it. But, there is the practical issue of getting it funded. That could stop the idea from happening. So, if you want it, get started finding funding and rallying support.