The second meeting of the new Content Improvement Group (CIG) was this week. The group has not yet been added to the SL User Groups list. You’ll find it in the SL Wiki as: Content Improvement.
Geenz Spad is known in viewer development circles. Geenz facilitates the CIG meeting. With a viewer development background it is probably not surprising that the first topic of discussion was outstanding JIRA items. It isn’t THAT boring…
Using the Linden Lab™ viewers, both main release and development, I kept crashing. I finally switched to the Exodus Development version and managed to stay connected (Exodus is fast). What a pain having to relog 3 times.
The highlights from the meeting follow. There is more on the Deformer debates.
VWR-27189 – Cloth simulation. This is a feature request by Paola Taube dated 18/Oct/11.
Cloth Simulation in 3D games is a challenge. Having cloth simulation in a game means a flag can ripple in the wind or a dress can swish and swirl as one walks. We sort of fake it with Flexies.
My experience with cloth simulation in a technical sense is mostly with Blender. I know running a cloth simulation in Blender 2.63a brings my computer to its knees and is slow. So, I tend to think it is something unlikely to be added to Second Life™ any time soon. But, there are apparently more efficient ways of handling cloth simulation than what Blender uses. So, I may be overly pessimistic.
Falcon Linden has voiced an opinion that cloth simulation is something he would like to add. Gwyneth Llewelyn has commented in related JIRA’s about efficient cloth simulation algorithms.
VWR-29210 – Cannot put more than one object into the contents of an object. This is a bug reported by Loki Eliot dated 01/Jul/12.
He is seeing it in the Development (220.127.116.110726), BETA (18.104.22.1680934), and Second Life 3.3.5 (260726) viewers. I’ve not been packing stuff into object contents so I have not noticed the problem. If you have, visit the JIRA and click watch.
The SL Viewer version 3.3.3 (260300) Jun 21 2012 does not have the bug.
SH-2103 – There is a gap between the shirt and pants layers – both shirt and undershirt seem to stop short of the waist seam. Reported by Siddean Munro dated 03/Oct/08.
There are images included with the JIRA, so you can see the problem being discussed.
Fixing this issue is going to require changes to the SL Avatar. So, I don’t find it surprising this has gone unfixed for some time. The problem should be kept in mind as we discuss other possible changes to the avatar.
Geenz makes some interesting comments about avatar changes. We know that for now changes to the avatar are considered a ‘no-no’ for now. But, Geenz says the Lindens are working on some ‘behind the scenes’ changes that will allow avatar changes somewhere down the road. That is optimistic.
I wrote most of the article The Consensus from information from the Content Improvement meeting. If you make mesh clothes, please take the survey poll at the end of the article.
Geenz describes Morth Targets as: “…a morph target is a special shape that defines how a mesh should “deform” when a certain parameter is increased or decreased.” I have some information on Shape Keys, Blender label for morph targets, in my tutorial Working with #SL Male Shape.
Geenz further describes morph targets, “Now, to some degree this could be seen as yet another mesh deformer alternative; although it’s one that would in theory be more flexible than cBones and Qarl’s deformer, and would put content creators in complete control over how a mesh ‘deforms’.”
cBones are the deform solution explained by Redpoly Inventor at a recent Content & Mesh Group meeting. I usually call it the Collision Bones or Alternate Deformer. I tend to refer to Karl’s, or Qarl, solution as the Mesh Deformer or just Deformer. Now we have Morph Targets as a way to deform mesh clothes. So, we do have 3 possibilities for making mesh clothes fit as we want.
Other games and 3D systems use mesh clothes on mesh avatars. But, Second Life is unique in that it needs a generic system that can deal with whatever we throw at it. AFAIK, there is no other 3D environment with that criterion.
In Second Life we have a unique mix of design freedom and limitations. SL users are learning to deal with the limits newly revealed by mesh.
It has been relatively easy making system clothes, those that use the system shirt, skirt, etc. One learns about the clothing templates and the problems associated with them. Most users can skip the idea that the Chip Midnight templates are actually UVMaps. There was no need to learn that aspect of 3D modeling. The Lab has effectively made using UVMaps simple.
Now that we are making mesh clothes we are running into additional problems and we have to deal with UVMaps in a more complex way. We are getting into the more professional aspects of 3D modeling.
Maxwell Graf brought up the issue of the level of difficulty required to make mesh clothes. The perception of SL that I have is that it is an excellent pathway into learning 3D modeling. Coming from the Myst MMOG a number of people wanted to make Myst like worlds. For many of us Second Life became our path for learning 3D modeling. So, keeping things simple and being able to stage learning so that we can get gratification at various level of learning by being able to make something usable is an important consideration.
At the same time, removing limits that frustrate professional 3D modelers is important too. Saying it another way, the learning process needs to work well as one moves from the most basic to the most advanced. Allowing people the freedom to fill the niche they are comfortable in will probably help player retention.
Looking at the level of difficulty of each of the ways we are considering, for clothes to fit, I see the following.
Karl’s Deformer is about the easiest. His Deformer adds a couple of check boxes. Otherwise all the things like modeling the clothes and weight painting them remains unchanged. We can consider mesh modeling and weighting a basic requirement of any mesh we hope to animate.
Collision Bones as explained by Redpoly will require that we learn more about animation armatures, vertex groups and the relations between the various aspects of the SL avatar. I am still wondering how animations and weighting is supposed to work with this process. It looks like it is going to require some add-on software for Blender. I think I can work around the need to add software, but it will likely mean learning a confusing dance to use the base and collision armature setups. So, I see it as more complex while adding more control. There is also the concern that is will require additional weight painting.
Morph Targets or Shape Keys are an interesting alternative. The way shape keys are made is going to require learning more about 3D modeling of the mesh. I suppose one could consider that work is shifted from the additional weight painting required by cBones to additional the mesh modeling required by shape keys.
Everyone likes to have control. I’m not as OCD as Christian Grey (50 Shades of Grey) is about control. But, in 3D modeling control is a good thing.
Karl’s Deformer provides the least control and the least effort on our part to use it. I think it is much more in line with the basic Lab philosophy of KISS, Keep It Simple Sister.
Collision Bones provide more control. It is hard to say how much it can add because we don’t know how it would be implemented. But, I think control will be very similar to what we have now. I think this option extends the style of control we have now to more bones. But, that is just my opinion from what I know.
Morph Targets as best I can tell now will add the most control. I don’t see them being any more complex than the others as far as learning. Using the feature is going to be work, but probably not that much more than we do now with multiple sizes. It will require more learning and that is counter to the KISS philosophy. The level of control seems amazing. One can control the deformation of each vertex in a garment.
If we have Karl’s Deformer and Morph Targets I think we would have the best possible solution. One could learn to make mesh clothes using just the Deformer as the entry level to mesh clothes. Advanced designers that so choose could go on to learn about morph targets/shape keys. So, I’m a do them both kind of girl…
I would like to see the Deformer completed as soon as possible.
Be sure you read the comments on The Consensus. There are several well made comments that disagree with me and see things bit differently. Also vote in the poll that is at the end of that article.
Morph Targets/Shape Keys are a future possibility, as far as I am concerned. There are many things regarding the SL avatar that need to be fixed, changed, and enabled to allow us to move to a more advanced 3D world. I think any implementation of Morph Targets is a long ways down the road. We are literally just starting to prepare the proposal to submit to the Lab for consideration. The Lab is going to take the proposal and figure out how they can fit into a comprehensive plan that works with their other development goals.
I suspect many will consider the process as too slow. However, good planning always reduces the time to goal. Plus we have many aspects of the Avatar that we, the users, want improved. The Deformer is to some extent and independent improvement. I see Morph Targets as more of a PART of a comprehensive avatar fix. So, we will have to deal with taking time to plan.
The more I think about morph targets, the more I like the idea.
I still think it is essential to have a system which can autogenerate them (such as what Qarl’s deformer is essentially doing by proxy through the avatar mesh), to avoid forcing every creator to go through the steps all the time and introducing another workload- and learning barrier.
But, apart from giving us more – fairly well-understood and standardized – ways to work with content, it could also be a way to break the stalemate over the avatar upgrade. If clothes are keyed to the morph targets rather than the avatar skin, you can switch the avatar mesh underneath it with no problems. It would even be possible to auto-convert existing system layers to a mesh piece matching the old avatar topology, and wear that.
Pingback: Making a Difference | MetaReality Podcast
I’m not sure cloth simulation as such is a very good idea. The demonstrations that are there focus on a very small number of avatars for decent simulation resolution. I suspect the performance of a mixed CPU/GPU simulation and pure CPU simulation to be similar across high end machines.
If you’re curious, get Nurien demo and see how it performs. It’s GPU-powered. It works just fine when there’s only one character in the simulation on my computer. The problem with cloth simulation is not to much the dynamics, with verlet integration that works quite easily. It’s collision detection and coupling the dynamics to the collision detection.
It’s curious though what Cloak Works can accomplish. They are supposed to be shipping next update of Rift with that cloth simulation technology, but so far all videos i have found, i could not spot any cloth simulation.
Flexies are heavyweights, and they’re clumsy, i’d very much like them to go. But what kind of suitable, low-end, large-scale capable replacement can we offer for them? Jigglebone systems have worked very well on older games, on older game plattforms, and provide rather decent cloth, boobs, hair, you name it. They are a system of bones, bound to avatar skeleton, which are simulated with springs and constraints. I think they could fit very well into re-imagining the avatar system based on mesh, but it does need some careful thinking how exactly. Unless the constraint system is made really very very complex, of course this system sets limits to what designers can do, but it would not exclude owners of ordinary computers and work well performance wise, and can perhaps be extended to more constraints later.