#SL Avatar Changes ‘in progress’

If you are working with mesh you probably understand some of the problems the Second Life® avatar has. If you make clothes for Second Life Avatars you are also likely aware of the short comings of the Second Life Avatar. For all the time Second Life has existed users have wanted better avatars. When SL Avatar 2.0 is released that will change… we’ll want better avatars. Oh,  wait! That’s not a change.

Another gotcha at the max breast size setting

We always want something better and as technology improves we get it. It takes better computers to run SL than it did when SL was created. In 9 years (going by SL birthdays) technology performance has doubled about 6 times, or 2^6th = 64 times… don’t see it in your frame rates? Neither do I.

For some time when I see a Linden asked about when we might see an Avatar 2.0, I hear them give evasive answers. Not maliciously, they either don’t know or are NDA’d. The best non-committal answer we get is that it is not a near term project. I think the Lindens would love to give us a new avatar. The problem is the literal RL millions of dollars of content such a change would break.

The Gotcha

Those that work with making clothes may not realize the clothes templates are UVMaps. A template/UVMap is part of a mesh. It tells the system how to stick the clothes textures on the avatar. It maps the texture to the mesh. If you change the mesh, you have to change the map and anything but minor changes means it breaks content.

It is possible to create a new avatar mesh and map it to the old UVMap. While that would allow the old textures to be used, it is far from ideal. For one thing, really good textures made for the current map take into account the foibles of the map. A new map would not have the same foibles, or what would be the point of a new map?

Second, if one is going to all the trouble to redo the avatar, would one really want to restrain it by making it support its legacy predecessor? So, I think Avatar 2.0 is not a near term, if ever, change.

In Process

In the SL Forum some peeps were talking about improving the avatar. At times I guess I’m a wet blanket. I pointed out that the Lindens are working on other projects than the Mesh Deformer, Qarl Fizz’s project, and avatar enhancements. Until they get this round of fixes and performance improvements handled we are unlikely to see any work on avatars.

Oz Linden popped in to tell us that specifically my statement is not true. The Mesh Deformer STORM-1716 and STORM-1800The vertex weights of the default character mesh could be better are in progress. Now that is and is not a surprise.

The Deformer project started in late 2011. We saw the last update to it in early February 2012. We saw the last Linden Deformer Project Viewer in early January. (Reference: Latest Deformer Project Viewer – notice the URL, it changes once you get to the page) I haven’t seen any movement since those two mile stones. Oz did say he was too busy to get the Deformer 0.2 merged into a current build of the LL viewer. (Reference: See March 12 in STORM-1716). That was a month ago. But, I would expect one to say the Deformer is in progress. The vertex thing is a surprise.

However, if they are too busy with other things to handle the Deformer merge and re-weighting the existing avatar vertices, I am guessing that something is ahead of those projects. I really doubt we will see work on the avatar and deformer until those things, whatever they are, are complete.

Vertex Weighting

I did find it surprising that avatar-vertex-weighting is an in-progress-project. It had also not occurred to me that the avatar was poorly weighted. I did notice how the stomach area distorts with even small movements. Other parts of the avatar are also poorly weighted. I noticed all those, but it didn’t click that was a vertex weight problem.

Changing the vertex weights will make the movement of the avatar render better. It can improve the appearance of the avatar and clothes when the avatar flexes.

It is also possible to make some minor changes to the avatar mesh and its UVMap without moving to a full Avatar 2.0 upgrade. Adding a few vertices or edge loops here and there and keeping a similar UVMap will allow improvement without breaking all the content. So, some improvements I haven’t considered are possible.

Consider the problem of making short-shorts and panties for the avatar. In the groin area there are some annoying distortions and texture stretches on the inside of the thighs. Improving the avatar mesh by adding some vertices and changing the UVMap by adding vertices could alleviate some of those problems. Technically this would break content. Textures that are distorted in the groin area would still distort but differently, which probably wouldn’t matter to most of us. But, new clothes could take advantage of the improvements and allow better panties and shorts with less distortion.

People are already playing with the improved avatar. As the avatar is built into the viewer, you can change it in your viewer. That changes how you see your avatar and everyone else’s avatar. It doesn’t change what anyone else sees.

Vertex Weights and the Deformer

Changing the weights of the avatar’s vertices will have an effect on the Deformer… or can… I guess that depends on how it is written. But, better avatar deformation will likely mean better deformation by the Deformer. The picture above shows the crumple that is a problem as breast size changes.

So, in some ways the two projects are tied together.

TPV Policy

Technically if you’re changing the viewer for your personal experience, it doesn’t matter to the Lab. When the TPV Dev’s start making that change, we may run into some Linden objections… or not. This is one of those fuzzy areas where nitpicking technical issues cloud the issues and probable outcome.

Summary

So, we may get some avatar enhancements in the near term… meaning months.

6 thoughts on “#SL Avatar Changes ‘in progress’

  1. It does seem that ‘in progress’ applies to things in the ‘nice idea, maybe one day’ category’ – and I know that at least one thing that was brought to the verge of implementation months ago was stopped and been left as ‘in progress’ .

    With all the recent and planned improvements in the appearance of SL it is a shame that the avatar mesh is likely to be one element that still makes SL look dated.

  2. Wow, another great post!

    I’d been reading your site for a while and I guess you could say I’ve been a “lurker.” Your articles are very technical– which, thankfully, I understand. As someone who has been experimenting a lot with mesh technology, I just kept nodding my head. I, too, have wondered why there aren’t more vertices in critical areas, such as the shoulders, knees, chest, and crotch. But we’re all learning, just as the folks at Linden Labs are.

    If Linden can add more vertex density and more sensible weighting, we could solve a lot of problems without straying too far into the “it-will-not-work-with-legacy-goods” domain. Just changing the vertex density won’t be enough, I don’t think, as they would still be subject to the inaccurate legacy weights.

    The weights that they use currently seem to be tied to the brute-force application of the “Skin” modifier (in 3DS Max parlance) rather than the newer “Physique” modifier– which allows for the weights to take into account muscle flexing and what could be described as the “density of the tissue” in a given area.

    I did a line of mesh products (full body suits) that used the default female avatar pieces as the base and, i suspect, if I went in and re-jiggered the vertex density as well as the WEIGHTING, I could achieve much better results– the upshot being that their would be less interpenetration with the underlying avatar.

    Normally interpenetation isn’t a big deal (many mesh products ask you to put a alpha mask over the area that is being covered), but my full-body suits are partially invisible, so that wasn’t an option for me.

    Thanks for making me think harder on solving this! I will need to adjust the default weights on the default avatar they’ve given us, producing far better results!

  3. I’m still waiting for an answer from a Linden for a project that has a deadline. That was 2 years ago and the project deadline about a week less than 2 years so when they say “changes are being worked on” I’m not confident. As far as performance goes I have been on Sl for about 3+ yrs and each time I upgrade my PC I get a huge increase in fps. Also the viewers especially TPV seem to get much faster.

  4. Pingback: Mesh Deformer update: release from Qarl, request from Oz | Living in the Modem World

  5. I do wonder if rigged Mesh has made changing the default Avatar more difficult.

    It depends on using the Avatar rigging–the internal structure and vertex weights and stuff–and changing any of those would break the Mesh content that used it.

    The sort of tweaks to mesh structure and UV mapping you describe, they would also involve the data applied by the Appearance controls. It’s a bigger project than you might think. Still, after so many years, the lack of change from the early days looks pretty bad. Does Linden Labs have anyone who can do that sort of job?

    I have a feeling that the release of importable Mesh was badly thought out, at the strategic level. Some of the problems associated with it might have happened anyway–the “bugs” in OpenGL for one–but in some ways the project felt like the tail wagging the dog.

    • I suspect changing the base avatar mesh should have little effect on the Mesh Deformer. It could improve or hinder how it well it works. But, the Deformer and Viewer are using an array of vertices. I doubt that it matters which array of vertices it is given. So, a new avatar is probably a minor issue for it.

      The Lab has a number of people that could make the change. Talent is not the hold back here. I suspect if the Lab were to allow a change they would have users provide an avatar or have a professional model one. So, the talent is available. The problem is in avoiding breaking existing content.

      Form our side of SL it can look like the project was poorly planned. But, we do not have access to all the criteria the mesh project, the time lines, nor even the complete plan. So, we cannot see what parts are down the road, off the table, or even the percent of the project that was included when they said ‘complete.’ They did make some mistakes. Some of those can be charged to BOTH poor planning on the Lab’s part AND poor input from the community. But, we have no objective way to critique the overall planning.

Leave a Reply

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