Metaverse Avatars

[Ro Gastel]

Thought i would comment here on avatars in SL and the future. Some
precautions for everyone to chew on.
Poly count may or may not be an issue for todays graphics cards. It
seems the SL's standards are a bit low when it comes to that and of
course we should not go over board on the counts We should try to do
something like normal maps to get better detail and keep poly count
low as others have mentioned. What i want to touch on is how things
work thus far in a client/server based Metaverse.
Opensim and SL have a basic avatar that everyone has that is loaded
onto there client, previously known as "ruth".
What you look like in SL is a morphed change from that basic avatar,
Male, taller, shorter, larger feet. This is done so that the basic
avatar that anyone sees is already "known" to your viewer and then the
actual avatar you see is some changed version of the base. This makes
avatar rendering fast and log into the "world" faster.
If we had to download a bunch of different avatars before looging all
the way in or worse yet you can't see them in-world until you have
gotten all there "bits" it could cause a lot of traffic. Imagine
having to download thousands of avatars to keep up every time you log
in:O.
So what the modelers of avatars want to try to do is build a better
base and have the same and better morphs that let you take that base
and make it personalized.
I think the only open metaverse provider that has universal avatar
storage is realexted but i am not sure how the handle getting that to
the client when it tries to log in. Anyone know?

Just wanted to put my 2 scents in

[Dahlia Trimble]

I think we really need to look at avatars as an application level concept. OpenSim is virtual world platform and can be used in several different types of applications. A *very large* portion of the OpenSim user community experiences it through the Linden viewer or some derivative. Avatars are defined in the Linden viewer and OpenSim is capable of storing and transmitting "appearance" data for these avatars, part of which are morph settings. Defining a "standard" avatar that differs from the one used in the LL viewer would deviate substantially from the vast majority of OpenSim users and applications to date. Alternative applications would need to extend OpenSim somehow via additional add on modules or some other means if other avatars are desired. Realxtend has added a new format of avatar to their viewer and they have modified OpenSim to be able to work with this avatar.

Idealist viewer in its current form allows avatars to be installed at run time by making the appropriate mesh and texture assets available and configuring the viewer. Given that Idealist is early in its development and that it's an open source project, this feature may evolve or change and will likely be application dependent. I see no reason why a particular application couldnt download an avatar from a server somewhere and allow multiple viewers to share the same view of multiple avatars which differ in design. If one spends any time in any of the content creator circles in Second Life, one might become of the opinion that this is a *very desirable* feature as people have to jump through many hoops to design alternative avatars and make them work within the constraints of the current Second Life environment.

Idealist is not based on the latest state of the art 3D technology... It's based on the Irrlicht engine which is an open source project and is not really up to speed with the latest advancements in the works at companies like Nvidia and ATI. There are many advances coming about in GPU design and within a few years we may see real time ray tracing of very high poly models with many features such as high speed client side physics processing. Avatar technology should improve also and I'd prefer to leave the door open to these improvements, rather than attempting to specify a "standard" avatar based on a snapshot of available technology at a given time. I think that if any standards come about, they may be in the form of a minimum skeleton layout for animation of bipedal avatars or even in the form of standard minimum animation sequence messages such as "walk" or "sit" or "fly". There can be additional criteria built upon these minimal messages but they would likely be more application dependent. Of course having a lowest common denominator of compatibility for assets such as animation data or UV mapped avatar textures would be a good thing, but I think a lot of careful planning/research/development/testing/user feedback needs to be involved before anything may be declared a standard.

Anyway, on a side note, I did a little testing with the makehuman avatar. I first attempted to import it into irrlicht with no luck. I then took a collada version of the avatar and imported it into blender, then exported it in blitz3d format and imported it into irrlicht. The poly count was above 36,000 (sorry I dont have the exact number at hand). After importing, the FPS of the irrlicht meshviewer application dropped from 70-80 to just above 40 on my machine (intel core 2 duo 2ghz, nvidia 8600M). I brought the avie into Idealist and the FPS dropped from 20 to around 12 on the same machine and using similar viewing conditions. This should be contrasted to the Sydney avatar which has around 900 polys. Granted these are 2 extremes in mesh complexity. The base female SL avatar has 3912 polys and seems to be much closer to Sydney in complexity than to the makehuman avie.

I've also edited the makehuman avie to remove a lot of the internal geometry such as the skeleton - there are bones inside it such as ribs, vertebra, a skull, etc. and I successfully reduced the poly count to around 9000 with no visible changes to the outside of the mesh. I was attempting to animate it but blender was responding slowly while I was rigging the mesh and it was just taking too much time.

Another issue with using the makehuman avie is the internal rigging functionality does not seem to survive import into blender, hence it was useless and would need to be re-rigged. I assume that if a program such as makehuman was desired as an avatar editor, then it's export format may need to be modified and the consumer program - Irrlicht and Idealist in this case - would need to be modified to make best use of it

 If some makehuman-like program is out there which is more compatible with the licensing goals of Idealist, similar criteria should be evaluated before considering it as a potential source of content.

[Tommi Laukkanen]

Is the SL avie license open enough for IdealistViewer? It starts to
sound that supporting SL avie in idealist for casual user needs in the
same way as SL prims are supported could be good course of action.
Dahlia have you studied the SL avie and its morph options. Do you
think we could implement in in reasonable amount of time?

Second we have the custom avatars which is pretty much supported
already through irrlicht engine.

Third there is the possibility of creating new morph avatar based on
MakeHuman or some other system, self made or not.

-tommi

[Dahlia Trimble]

I don't know what the license for the sl avie is, I think the sl viewer "artwork" is Creative Commons share alike or some such, but I don't really know the details, or if the avatars and morphs are included in the same licensing terms as the rest of the viewer artwork.