Breaking News! Fitted Mesh Project Viewer

[Medhue Simoni]

---

Gaia Clary wrote:

As a sidenote: I do not understand (and never understood in the past as well) why LL does not inform their customers beforehand. All what i had until yesterday was an informal note that i got about a year ago from Linden Labs. They asked us to better not give collision bone rigging out to the public because the lab could decide to break this approach at any time.

---

That is the core of what annoys me about the whole thing. I can see some things being a secret, but why was this? Why allow your community to complain and get upset, when you are working on it. Just say that you are working on it or looking at this or that option. It's like they are playing a game with us.

Even I find it difficult to be upset for too long over this. The goal was to have a solution for mesh clothing and now we have something. You gotta be happy for the clothing creators, and their customers. In the end, especially this issue, I don't really care what solution they go with, as long as we have something that works reasonably.

As far as Oz's statement at the end of the mesh deformer jira, I'll just point out that Oz get's paid to be annoyed by those emails. We do not.

All that said, I'm still waiting for LL to put out a revised TOS. So, this collision mesh announcement means absolutely nothing to me. I haven't uploaded anything in months, and I've been working on projects for other platforms. I have no intention of uploading anything until the TOS is reasonably reworked.

[Nalates Urriah]

One of the big problems with the Deformer was the base shape. For those making large and small avatars, that includes fat and skinny as well as short and tall, use of the default avatar shape as a base was unsatisfactory. That resulted in the ability to specify your base shape for the deformation. Thus the XML file shape exports became a way to specify a base shape for the Deformer to work from. But, understanding how that worked and how Deformer offsets worked in general was/is difficult.

Using custom base shapes was seen as an overall problem and complication by the Lab. They were concerned about how custom base shapes were going to be explained to new users in the market place and how they would figure out which clothes could be used with their shape.

While people could have worked out a solution, history shows us that many uninformed people sell in the MP. So, there would be well explained items and poorly explained items. Basically a confusing mess for novice SL users. Anything that adds confustion or adds significantly to the learning curve is a problem for the Lab and they avoid it. Fitted Mesh eliminates that problem, confusion, and learning curve.

[Sassy Romano]

You write authoritatively on behalf of LL. They didn't communicate with us, please cite where they made all those statements. I'm interested.

 

At present it's not a particularly difficult task to use Avastar to rig normally and then transfer weights to collision bones where they exist in the same place. I haven't tried doing the pecs yet but that doesn't strike me as trivial.

 

What we seem to have is something that is low effort for LL and nobody else.

 

Anyone have any tips for rigging workflow for the pecs other than 100% trial and error?

[Cathy Foil]

I just got up so my brain isn't fully functional yet so I may not be following you Nalates.

I had the good fortune of talking directly to Qarl about how the Mesh Deformer works.  I said to him that it was my understanding that the Mesh Deformer works by taking the positions of the default avatar's vertices and then using those vertices to influence the position of the closest vertices in the mesh clothing or object that the avatar was wearing.

He response was YES that was exactly how the mesh deformer works.  Now my assumption is that the Mesh Deformer gets its initial information about default avatar mesh and its vertex positions from the XML files just like the viewer does.

The base shape for the Mesh Deformer is the default avatar shape itself namely Ruth.  If it is unsatisfactory it is because the default avatar mesh and more importantly its Morphs and weighting are unsatisfactory and not esthetically pleasing.  The Mesh Deformer works fine.  My own personal experience is that the Mesh Deformer, even with the unsatisfactory base shape of the avatar mesh itself, works better than Collision Bones.

Now when I talked to Qarl there was talk about using a base shape other than the avatar's mesh.  This would be basically an invisible mesh you have on who's vertices, weighting and morphs would while being close to the default avatars shape, result in a smoother deformation of the mesh clothing by having the vertex be more evenly and smoothly distributed in the base shape mesh.

Now this new base shape would be a massive pain in the butt to set up.  Basically the same amount of work creating a whole new avatar and make all the morphs to match the morphs of the default avatar and weight it and adjust the weighting so that it moved well.  Only thing you wouldn't have to worry about would be the UVs since it be invisible.

Now mind you if LL came out with a new avatar with better morphs, weighting and vertex distribution it would make a much better base shape and thus the Mesh Deformer would work even better.

========================================================================================

OK my brain is working better now and I just re read your post Nalates.  You are not talking about what I am talking about at all.

How is the Collision Bones method going to avoid all that you say?  It is still going to stretch the clothing mesh to fit tall or short of fat or skinny.  You still make your clothes on the "Ruth" the default avatar shape and skeleton settings.

The only way around that I can see if someone wants to make the initial mesh avatar clothes on the size and shape of the extreme avatars you are talking about would be to get all the Morphs and skeleton positions from LL and basically recreate all the sliders in the editing appearance menu in the 3D graphics program you are creating your mesh clothing in.

Then you would set your avatar in your program to the size and shape you want create your mesh clothing and skin and weight it to the skeleton then change all the size and shape settings back to the default Ruth settings and then export your mesh as a DAE file.

Now I believe this is what Avastar does.  It would be simple to setup in Maya because LL made the avatar and morphs and everything in Maya in the first place and would be great if LL made them available if they haven't lost them by now.

The same thing could be done for Qarl's Mesh Deformer.  All that would be needed is a plug-in to simulate the Mesh Deformer in Maya and Blender.  Make your clothing mesh on Ruth's shape, skin and weight it, change the shape of the avatar using the sliders to tall or short or fat or skinny, adjust the vertex positions of your mesh clothing then change the sliders back to Ruth settings and export your clothing mesh as a DAE file.

Is LL going to do any of this for us?  I highly doubt it.

 

 

 

[Hada Sabra]

Mine are not symetrical either.

>>Hmm, so I downloaded "Fitted Mesh Skeleton.zip" and found that the legs of female mesh model are not completely symmetric.

[Pussycat Catnap]

---

Cathy Foil wrote:

Not all deformers are created equal nor does implementing both have to be a system resource nightmare.

---

Calculus is calculus.

A famous question on this issue was "what about the buttons" - and once you understand that question, you begin to see why a deformer is never viable for a good result.

 

As to resources - OSGrids have negligible concurrency. Millions of calculations per second in real time need to keep something deforming - can blow up as it scales up. Even if its a done once and so when I see your copy of the same shirt I have, I download it as a different object... even in that scenario the process of originally deforming it means those millions of calculations had to be run once - and multiply that up as concurrency goes up and its a lot more significant than the math of prim objects being warped along preset dials (itself already a major cause of lag in SL. Just ask lindens about the horrors SL suffers from the existence of the torus prim).

Deformers have a place in static art - where you can pose something away from a poping triangle, or take it back into 3D modeling and manually fix 'math errors'. But in real time animation - its a concept that was doomed from the inception.

If for no other reason than... "what about the buttons?" - which hasn't and doesn't even get to the lag issues I've just talked about.

 

[Medhue Simoni]

---

Nalates Urriah wrote:

One of the big problems with the Deformer was the base shape. For those making large and small avatars, that includes fat and skinny as well as short and tall, use of the default avatar shape as a base was unsatisfactory. That resulted in the ability to specify your base shape for the deformation. Thus the XML file shape exports became a way to specify a base shape for the Deformer to work from. But, understanding how that worked and how Deformer offsets worked in general was/is difficult.

Using custom base shapes was seen as an overall problem and complication by the Lab. They were concerned about how custom base shapes were going to be explained to new users in the market place and how they would figure out which clothes could be used with their shape.

While people could have worked out a solution, history shows us that many uninformed people sell in the MP. So, there would be well explained items and poorly explained items. Basically a confusing mess for novice SL users. Anything that adds confustion or adds significantly to the learning curve is a problem for the Lab and they avoid it. Fitted Mesh eliminates that problem, confusion, and learning curve.

---

I made clothing with the mesh deformer and am selling it in InWorldz. It works with almost every size female possible. If I really wanted, I could have made 3 versions, all using the default base, and it would have fit small children, and the other fit the biggest strongest, fattest avatar in SL, meaning every1 that uses the SL avatar. I just didn't feel the need to. I even showed the video of me using all the sliders. There were no problems with the deformer at all. Making the clothing and rigging it was simple as pie. No real extra work, but maybe tweaking a little at the extremes. If some1 got different results than I did, I'd chalk it up to inexperience, before blaming code, which works the same for every1. It's really all about weighting the clothing perfectly. Once you get a perfect weight balance, you can use that to copy to any shirt or pants. Oh, and I'm not a clothing designer.

I must also refute any claims that the custom based shapes had any validity at all. They were not needed and never were. I personally think Qarl conceded only because it really didn't matter and he knew it. Well, obviously he knew that, but if it quelled fears, then why not add it. The actual bottom line was that you could have used any shape as the default, as long as that is the default for every1. IMHO, adding the other sizes only gave credibility to irrational people who didn't understand anything about the deformer. I show in my own product why the whole concept was BS, because I could easily get any shape to work perfect with only weighted to the Default.

[Whirly Fizzle]

---

Yabusaka Loon wrote:

Hmm, so I downloaded "Fitted Mesh Skeleton.zip" and found that the legs of female mesh model are not completely symmetric.

Does anyone have fixed ver. of skinned female mesh?

Male mesh seems ok.

---

See FITMESH-7

Fixed files are attached to FITMESH-2

[Cathy Foil]

---

Pussycat Catnap wrote:

Calculus is calculus.

Cathy Foil response:  Again not all calculus equation or the way you use them are equal

A famous question on this issue was "what about the buttons" - and once you understand that question, you begin to see why a deformer is never viable for a good result.

 

Cathy Foil response:  I am not sure what you mean by "what about the buttons".  If you mean mesh buttons on a mesh that is deformed by the Mesh Deformer look stretched or squished when they are supposed to be rigid you will get deformations with the Collision Bones as well.

As to resources - OSGrids have negligible concurrency. Millions of calculations per second in real time need to keep something deforming - can blow up as it scales up.

 

Cathy Foil response:  Yes but this is also true with Collision Bones that are set to avatar physics.  The bones as they bounce will continually deform the mesh weighted to them.  Anytime your avatar moves in 3D space calculations have to be made as the vertices change position.

Even if its a done once and so when I see your copy of the same shirt I have, I download it as a different object... even in that scenario the process of originally deforming it means those millions of calculations had to be run once - and multiply that up as concurrency goes up and its a lot more significant than the math of prim objects being warped along preset dials (itself already a major cause of lag in SL. Just ask lindens about the horrors SL suffers from the existence of the torus prim).

Cathy Foil response:  Again this is also true for Collision Bone method.  As the avatar wearing the mesh changes it appearance settings the mesh is deformed to fit requiring calculations to be made.  Now the Collision Bones calculations are probably less demanding than the Mesh Deformer but the difference is insignificant.  If there had been a real issue with calculations being made by the Mesh Deformer causing lag it would have been brought up in the two years of testing and shown to be a real problem but it didn't.

Deformers have a place in static art - where you can pose something away from a poping triangle, or take it back into 3D modeling and manually fix 'math errors'. But in real time animation - its a concept that was doomed from the inception.

Cathy Foil response:  That is your opinion and you are certainly entitled to it.  However the concept wasn't doomed from inception because the Mesh Deformer worked and works well.

If for no other reason than... "what about the buttons?" - which hasn't and doesn't even get to the lag issues I've just talked about.

 

Cathy Foil response:  Again with the buttons!  No system is ever going to be a perfect solution.  Not the Mesh Deformer not the Collision Bone/Liquid mesh.  There are possible solutions to the button issue.  Perhaps a texture map or material could be designated as some sort of marker so that it will only be deformed  by closest avatar single mesh vertex.   Meaning all the vertices of that face with that texture map or material applied to it are only affected by the closest avatar mesh vertex.  This would give the result of the buttons appearing rigid.  Something like this could have been implemented in the Mesh Deformer 2.0.  There are probably many many ways in which areas or groups of vertices would be designated in such a way.

As far as lag is concern you are worried about the little flame on the candle that is the Mesh Deformer all while the rest of the avatar is a burning bonfire with multiple objects attached running ridiculous amounts of scripts covered with a multitude of 1024 by 1024 textures and tens of thousands of vertices from prims and sculpties and yes even mesh.  That is the lag we should be concerned with.  Who's bright idea was it to allow basically unlimited objects to attach to each attachment point?

---

 

[Yabusaka Loon]

Thank you for the info!

[Medhue Simoni]

---

Pussycat Catnap wrote:

---

A famous question on this issue was "what about the buttons" - and once you understand that question, you begin to see why a deformer is never viable for a good result.

 

 

---

This is hilarious! Really it is. The solution to the, "what about the buttons" is to make the buttons separate. Seriously, was that all that difficult to figure out? As Cathy pointed out, collision bones won't be any different.

The rest Cathy covered pretty good, so I won't go into it.

[Syntella Rhapsody]

If the deformer did not work for you, maybe the problem is not the deformer, it's the mesh you made. As you can clearly see in Medhue's video the deformer works exactly as expected. I had the same results as Medhue. It worked fine for me. I did tests of making mesh that had a vertce layout different from the avatar and making clothes that were cut from the avatar. The clothing cut from the avatar always worked better because it is the exact vertice layout of the avatar. As long as I made sure the vertice layout mimicked the avatar in certain problem areas I had good results. Lets lay this out plain and simple. The avatar DEFORMS to make different shapes. When you use the shape sliders the avatar is DEFORMING. The "Deformer" works on the same formula as the avatar and will work exactly how the avatar does. The mesh that is deforming along with the avatar will go short, tall, fat, skinny, pregnant, LARGE breasts, small breasts, big butt etc ect ect. Because the avatar will deform to all those shapes. Give us BOTH! Let the creator make the choice on how they want to work with their mesh and creations, whether they want to use the deformer or collision bones. It's like forcing me to use Maya rather than Blender or visa versa. We all have our own preferences and work flow. LET US USE WHAT WE WANT! If you want to use collision bones, more power to you, create until your little hearts content, but don't refuse me the same respect. I helped pay for the deformer to be developed as did many residents. IT WORKS!

My only guess on why the Deformer was never implemented was purely political by LL, and if someone other than Qarl had developed the deformer, it would have been implemented a year ago.

Direct quote from the blog.

"Neither approach completely eliminates the occasional need for an alpha clothing layer to prevent small parts of the avatar skin from appearing through garments, but both work quite well at resizing garments so that they fit the avatar and move naturally with it."

Both work quite well.

Goodnight.

[Gaia Clary]

---

This is hilarious! Really it is. The solution to the, "what about the buttons" is to make the buttons separate. Seriously, was that all that difficult to figure out? As Cathy pointed out, collision bones won't be any different.

The rest Cathy covered pretty good, so I won't go into it.

---

Wouldn't the solution to this be to give all vertices of the buttons the same weight ? Then the buttons would scale but not stretch.

With the Mesh deformer you would have to make the buttons separate and static i guess ? But then wouldn't you get really ugly issues with garment animating due to avatar movement while buttons stay at their attachment place ?

[Cathy Foil]

---

Gaia Clary wrote:

Wouldn't the solution to this be to give all vertices of the buttons the same weight ? Then the buttons would scale but not stretch.

With the Mesh deformer you would have to make the buttons separate and static i guess ? But then wouldn't you get really ugly issues with garment animating due to avatar movement while buttons stay at their attachment place ?

---

Collision bones, depending on which slider you are using in the Editing Appearance menu, don't scale equally in all three axis of XYZ so you would get stretching of buttons a good amount of time especially on the extreme settings which LL is concerned about. 

[Medhue Simoni]

---

Gaia Clary wrote:

---

This is hilarious! Really it is. The solution to the, "what about the buttons" is to make the buttons separate. Seriously, was that all that difficult to figure out? As Cathy pointed out, collision bones won't be any different.

The rest Cathy covered pretty good, so I won't go into it.

---

Wouldn't the solution to this be to give all vertices of the buttons the same weight ? Then the buttons would scale but not stretch.

With the Mesh deformer you would have to make the buttons separate and static i guess ? But then wouldn't you get really ugly issues with garment animating due to avatar movement while buttons stay at their attachment place ?

---

You'd have to attach the buttons to separate attachment points, this way each button moves with the bones too. Yeah, it's a mess for the users to set up on a unique avatar, but I think that would be the only way for it to work well. Of course, clothing is not my market, so I'm sure others can think of better ways. If it was up to me, I'd avoid buttons, lol.

[Allison Lynagh]

I find this all remarkably interesting as a mesh clothing creator. However, I SCOURED the net for any help on how to accomplish this task. I create my meshes using blender, and I have avastar as well, though I still prefer using my method versus the avastar setup. That said, I find it awesome that Gaia has provided the skeleton and collisions and everything else, but disappointing that there isn't really any explaination on how to use this updated skeleton (The rotation alone throws me for a loop considering my 'front' view is really a side view, getting used to a forward facing avatar on the 'correct' axis is hard enough). I found some things that sort of suggest how to accomplish this task, but every explaination leads me to ask more questions as each attempt at a simple recreation has failed miserably. Any guidance on how to implement this in blender 2.69 would be a godsend!!!

[Sassy Romano]

The information has been around on the Machinimatrix support site for Avastar for a long time actually but it's not always easy to find.  This doesn't specify the few new bones but it's all the same and Gaia plans to do new tutorials.  She halted because LL made such a fuss about "don't use these bones because we don't guarantee their support in future". 

http://blog.machinimatrix.org/avastar/reference/use-of-collision-bones/

Video tutorial, the Collision part starts at 14:15

[Cathy Foil]

I don't know Blender myself but I can tell you how to in Maya.  The principles are the same you would just need to adapt them to Blender.

Basically all you have to do is weight all the vertices of your mesh to the Collision Bones.  The DAE file has to contain the Collision Bones you use and ALL of the normal bones as well in order for the DAE file to upload correctly.

So in Blender I suppose you would weight your mesh to all the bones in the skeleton, normal bones and collision bones.

The tricky part is adjusting the weights so that your mesh not only moves well when the joints bend but also scales well as the collision bones change their scales in SL when someone edits their appearance using the sliders.

The arms and legs collision bones are offset from the normal bones so when the normal joints bend your mesh will not move as you would normally expect.  Because now you have to adjust for the offset and try and get the vertices to scale smoothly it will take you 2 to 4 times longer to adjust the weights of your mesh.  It is easier with looser fitting mesh and harder with tight fitting mesh to get good results.

All your mesh that you have already uploaded to SL will have to be rigged again from scratch and the weights adjusted as I have described.  That is of course if you want your clothes to work with the deformer LL has chosen.  I am sure most clothing designer will want to do this in order to stay competitive.

Now had LL chosen the Mesh Deformer all you would have to have done was weight and adjust your mesh like you have always done to the normal skeleton bones and then when uploading your mesh pick two options in order to have your mesh work with the Mesh Deformer.

This would have meant that any DAE file you already have from previous mesh could be uploaded in just a minute or two.  

Thanks to the Collision Bone method having been chosen, if it used to take you an hour to adjust your mesh weight it will now take you 2 to 4 hours.  Instead of taking just a few minutes to upload each of your old DAE files it will take you 2 to 4 hours to get each one ready.  Isn't that wonderful! :matte-motes-sour:

[wesleytron]

How well would this system (as compared to the mesh deformer) work for making and applying to mesh avatars? Which would be better at adjusting, say, the shape of female human mesh avatar using the standard sliders?

[Sassy Romano]

---

wesleytron wrote:

How well would this system (as compared to the mesh deformer) work for making and applying to mesh avatars? Which would be better at adjusting, say, the shape of female human mesh avatar using the standard sliders?

---

Since there are no facial collision bones other than ONE for the head, my understanding is that this method won't work well compared with the deformer by nearest vertex when wanting to deform a human to the LL provided model.  I could be wrong but that's my understanding.

In reality, would I want to deform a replacement avatar head to the LL morph sliders?  Probably not since I would surely want an improved model, not one that will be mushed to something nearby.  Personally, I haven't tried that and have no desire to.

 

[Theresa Tennyson]

---

wesleytron wrote:

How well would this system (as compared to the mesh deformer) work for making and applying to mesh avatars? Which would be better at adjusting, say, the shape of female human mesh avatar using the standard sliders?

---

With this system I believe you could replace the entire avatar from the neck down and have a much smoother mesh that responds to the existing shape sliders and could re-use existing shapes.. You could keep  the head of the current avatar mesh to retain facial expressions, and if there were texture appliers for the new mesh body that were compatible with the old skin texture layout you could use existing skins for it too.

 

The old avatar mesh is being used less and less lately anyway - taking account mesh feet/hands/breasts and mesh clothing that hides it with alphas, you often don't see any of it besides the head anyway.

With the deformer you'd basically need a 1:1 match between the new and old vertex locations and that would mean you'd have all the distortions of the old mesh.

I've seen the deformer in action both when it was in beta and on Inworldz - it works well enough with meshes that are supposed to be skin-tight but if the avatar mesh gets pointy or flat in an area the clothing will have the same flaws. If you've got something like a jacket that stands away from the body and the shape you put it on has major differences from the default avatar shape it  can end up looking like rumpled tinfoil.

[Innula Zenovka]

Forgive me if this has been raised, but if I download the project viewer, are there any items I can test it with?   Should existing Liquid Mesh items work with it, for example?

[Sassy Romano]

---

Innula Zenovka wrote:

Forgive me if this has been raised, but if I download the project viewer, are there any items I can test it with?   Should existing Liquid Mesh items work with it, for example?

---

Anything already rigged to collision bones will work but then they'll work in the standard viewer anyway.  You'd need something rigged to the new bones in order to see it function differently.

[Cathy Foil]

---

wesleytron wrote:

How well would this system (as compared to the mesh deformer) work for making and applying to mesh avatars? Which would be better at adjusting, say, the shape of female human mesh avatar using the standard sliders?

---

The Collision Bones method LL has decided to go with does not work well when it comes to making mesh avatars or replacement avatar body parts.

The Collision Bones method is more like having balloons attached to the bones.  You inflate or deflate the balloons and any mesh that is covering the balloons gets stretched bigger or smaller.  This is not how the avatar shapes are changed.  They are done with what are called "Morphs". 

No matter how you weight the vertices to the Collision Bones the vertices will not follow the shape of the avatar's Morphs.  This is fine for loose fitting clothes but the tighter the mesh clothing the worse the results will be.  This means making replacement body parts will not attach perfectly.  There will be gaps that grow bigger as you move your shape away from the default avatar shape.

If Collision Bones are added to the face this would allow for limited facial expression.  The more bones that are added the bigger the variety of facial expression that can be made though it is considered to be very primitive way of doing it.  The lips alone would need 16 bones arranged in a circle pattern. One for the jaw, One for each eyelid upper and lower. Two or three bones for the cheeks. Three bones for each eyebrow. Two for the corners of each eye. I think that makes a total of 37 bones.

For the hands LL would have to add 4 bones per finger and 3 bones per thumb so that would be 46 additional bones and probably 10 bones for the toes which would be needed to satisfy shoe designers.  Plus two more bones so each foot would have a heal bone and ball of the foot bone.  

So to make the Collision Bone usable for creating complete new mesh avatars only and not replacement body parts we are looking at 107 additional Collision Bones.  

For those who are preferring the Collision Bones method over the Mesh Deformer because of misperceived performance issues thinking collision bones will cause less lag just consider the possible lag caused by everyone all residents having to have 107 additional collision bones.

Let's look at the Mesh Deformer when it comes to creating full body custom and replacement parts.  Nothing needs to be added to the Mesh Deformer to make it work with full body or partial body mesh replacements.  Facial expressions, hands and fingers all work.   All the sliders for the avatar work.

For replacement body parts as long as the vertices where the body part attaches to the avatar are the same number and in the same exact places and weight of the default avatar mesh it matches perfectly no gap.  I know I have tested this.

Just one rule to remember.  The farther away your vertices are from the positions and weights of the vertices in the default avatar the less likely the mesh will conform to the avatar true shape.  So crucial areas like the edges of lips and eyelids and fingers you will want to stay closer to the original vertices of the default avatar.  Other than that the sky is the limit.

I had made a suggestion that a special attachment point be made, for the default avatar, that when a mesh is attached it would texture the mesh with all same texture layers that the default avatar was currently wearing except for the alpha layer since that layer is need to make the default avatar invisible.  A second Alpha layer for mesh clothing would also need to be added. 

P.S. Any new improved default avatar LL created the Mesh Deformer would work even better.  Don't worry LL will never get rid of the original default avatar just a second or third new default avatar would be added as additional options.

[mtwtfss71]

I can attest to the BELLY collision bone creating a mess once the belly slider goes above about 14. I hope more work is done on that one in particular.

 

Also, are people successfully uploading mesh clothing using the collision bones into the testing viewer? I'm getting an error even though according to blender, my mesh is fine.

 

[edited for typo]