MESH DEFORMER PROJECT NEEDS YOUR VOICE

[Ashasekayi Ra]

RE: The clipping issue

I think part of the problem is unrealistic expectations of what the mesh deformer and/or Avatar 2.0 will be able to do  Even if we get a perfect Avatar 2.0 (which I haven't seen any Linden officially agree will be created), there will never be a way to achieve zero clipping for every single clothing item made by various creators with varying topology. The issue of clipping gets worse when you include layering clothing with different topology created by myriad people. Let's not even talk about how impossible it would be to have zero clipping under those circumstamces when you add in animations. The use of alpha masks with clothing and the default SL avatar won't vanish. Although, the deformer will diminish the need for them in some use cases.

[Rahkis Andel]

I guess you never made it to page 3 because that's exactly what I said.

[Codewarrior Congrejo]

Hey Nalates,

So, while I would very much like to see this version released and other things fixed later... that may create more problems. I feel wanting this version is a bit self centered of me.

i totally understand. And it's always good to see new features released. Stable or not, in the end all of them add to a better experience,  more variety and choices, so nothing wrong with that

So while I can't say I'm not sure if I want the Deformer as it is released, I do, I am willing to wait for a more comprehensive and well thought out decision.

yeah that is kinda exactly the point where i am finding myself at.
And i am sure both sides could find a middle way on what is needed to accept it as 'release-able'.

[Cathy Foil]

I am with you White Rabbit!

LL just needs to release the Mesh Deformer now!

I put in several weeks working and testing the Mesh Deformer well over six months ago.

It worked fine back then it works fine right now.

Oz I don't think people have lost interest in it.

The reason less people seem interested in it is because we are tired of waiting for LL to impliment it.

For how long and how many times can we test something that works?

Please release the Mesh Deforemer. :matte-motes-kiss:

 

[Coby Foden]

---

Ashasekayi Ra wrote:

RE: The clipping issue

I think part of the problem is unrealistic expectations of what the mesh deformer and/or Avatar 2.0 will be able to do  Even if we get a perfect Avatar 2.0 (which I haven't seen any Linden officially agree will be created), there will never be a way to achieve zero clipping for every single clothing item made by various creators with varying topology.

---

I found this video very interesting:

http://3d-studio-max.wonderhowto.com/how-to/animate-clothing-without-your-avatar-skin-penetrating-outer-mesh-379834/

Now, if we had better avatar mesh, then that mesh could be used as a base for clothing items, by copying it and expanding it a bit.  Then the avatar and clothing item would have the same exact topology.  It could well be so that in a situation like this the deformer would do near perfect and amazingly excellent job.

 

[Innula Zenovka]

---

WhiteRabbit0 wrote:

Also Oz has repeatedly said this project is low priority because he thought it was evident that the community didn't care about it.  I'm trying to show otherwise, again.

---

Has he repeatedly said this?  Where and when?     I'm not disputing it, since while I do try to keep up on the jira and follow Nalates' accounts of meetings (and chat with friends involved in the project) I haven't been following developments as closely as I might, so he might well have said it without my noticing.    

If you could point me to a couple of places where he says this, I would be most grateful.

 

[Rahkis Andel]

---

Coby Foden wrote:

---

Ashasekayi Ra wrote:

RE: The clipping issue

I think part of the problem is unrealistic expectations of what the mesh deformer and/or Avatar 2.0 will be able to do  Even if we get a perfect Avatar 2.0 (which I haven't seen any Linden officially agree will be created), there will never be a way to achieve zero clipping for every single clothing item made by various creators with varying topology.

---

I found this video very interesting:

http://3d-studio-max.wonderhowto.com/how-to/animate-clothing-without-your-avatar-skin-penetrating-outer-mesh-379834/

Now, if we had better avatar mesh, then that mesh could be used as a base for clothing items, by copying it and expanding it a bit.  Then the avatar and clothing item would have the same exact topology.  It could well be so that in a situation like this the deformer would do near perfect and amazingly excellent job.

 

---

You already can:

Funnily enough, when you make clothes this way you get near perfect results (Perfect, as in clothes that deform exactly as the avatar does, thereby not clipping). A better avatar would look better, yes (particularly if it were weighted more carefully), but it would not perform better with layers of clothes made by different artists.

If creators think Avatar 2.0 is going to make their clothes magically never clip, I've got some bad news for them: The reality is that either their technique is poor or they don't grasp that some clipping is inevitable and must simply be covered up.

And you know, thats unfortunately the impression I get from some people...

[Codewarrior Congrejo]

Very valid point, and just supports what i was trying to explain in my topology tutorials. 'stick close to the avatar's topology' and edgeflow in the areas that are being influenced the most within animations. (which is a good advice no matter how the quality of the underlying model is. 

So even if we get avatar 2.0 and people still keep creating those chess-pattern meshes or wild topology,  they will still have avatar flesh peaking through. That is nothing the deformer or anything else can or could fix , and lays within the responsability of the clothing and attachment creators : )

[Coby Foden]

---

Codewarrior Congrejo wrote:

So even if we get avatar 2.0 and people still keep creating those chess-pattern meshes or wild topology,  they will still have avatar flesh peaking through.

---

If we had better avatar mesh, with good topology, added vertices on places where the current avatar has very low vertex count, then a lot of good clothing items (like pants, tops) could be created by the simple method shown in the video.  Copy parts of the avatar mesh, expand, copy weights from avatar mesh and texture.  Done.  Then even relative beginners in mesh would be able to create well working mesh clothes fairly easily.  No need to create wild topology.

With current avatar mesh that method could be used too.  But the results would not be anything very great due to the poor avatar mesh.  The crotch area especially is very problematic as it does not have any resemblance to the real human crotch at all - it's some strange canyon and the vertices there are light years apart from each other. :smileyfrustrated:  :smileytongue:

[Rahkis Andel]

If you really examine the edge flow of the current avatars, you'll notice that asides from some optimizations with old technology in mind, it was probably all quads before being triangulated. The mesh has a pretty good distribution of edges too. The only time it turns into a canyon is where the weights were sloppily executed.

Besides, any avatar 2.0 will not be significantly more high-poly than the one we have now (~7000 tris).

Having a crotch that resembles a human one will -not- make clothing generation easier anyway. Unless you are making skin tight clothes, that copy-paste method just isn't very optimal. Most practical, well made clothing will mimic the avatar's edge distribution, but not perfectly copy it. There may be some clipping in crazy animations, but that is what alphas are for.

To repeat what I said before, all a better avatar is going give us is a better looking avatar, not better looking clothing.

 

[Coby Foden]

Better avatar mesh surely would make it easier to design skinny clothes by the copy method.  Making skinny clothes for the avatar from scratch is not very easy for those not very experienced in mesh clothes making.

Or is there perhaps some rule "thou shalt not make mesh clothes by any easy method - it will be frowned upon!"? :smileyindifferent:

For loose clothing the clipping does not happen so easily, if at all, as for the tight ones, thus creating loose clothing from scratch is easier - in that respect.

A bit of "whining" outside of the subject: :matte-motes-big-grin:  :smileytongue:
I wonder, why is it so that Second Life has had many good excellent enhancement towards graphics quality, but updating the very ancient avatar mesh, for some strange reason, appears to be a big no?  How long we will have that mesh?  For all eternity perhaps?  What is the real underlying reason that there are opinions that there is no need to update it?  On the other hand there are also opinions that it is indeed very outdated and needs badly an update.

 

[Codewarrior Congrejo]

Heya coby,

of course a nicer layed out topology will make it more 'conclusive' when you want to do the copy avatar faces - to - create- clothing method.

You might have misunderstood me a bit here : I was just making aware that also with a new avatar you still have to take it's topology and edgeflow into account when you want to avoid clipping or misbehavior while being in an animation.


(the closer you stay to the same edges and even vertices in the important spots - no matter the quality of the model - the clothing item will of course work better with it). Maybe you got my point now =)

And no there is no such rule.. except the one where i would like to slap onto some fingers when seeing some weird topology or extremely dense meshes.. but that is something i most likely would have to discuss with an anger management trainer >.> *laughs*

But apart from that I stick with Rahkis and repeat my own statements I made in the past. The avatar isn't too bad of a model. I've seen others from games where you'd say that looks awesome.. and has even worse topology. It's always a matter of creating something that is optimized for a certain usage and engine. And most games actually 'trick' the eye by using very limited animations that do not reveal the issues or critical parts on their models, or by using different models for different extreme animations. Where they spare i.e.. more polygons on the legs and put higher density on the arms and so on. All of this to keep the model itself always at the required amount of polygons for the use in a certain engine.
Meanwhiles the Avatar in secondlife has to be an all-round usable figure, no matter the extreme poses users could create, they can't fall back on tricks and tactics other games or engines would use.

About your -quote- "whining" (let's better call it your personal wishing - that sounds nicer : ))

I think it had been stated very good on another spot, why the avatar isn't something so easy to replace.
Second Life wants to keep the 'shared experience' and downwards compatibility.
And unlike 'regular' games where you just install an all-new-version, have new content - secondlife is nothing like this.

The problem lays within different things: 
New avatar = new UV layout. you can 'try' of course to match them a bit up again but generally the rule stays like this : no old clothing items, skins, etc will work anymore with the new one.

Additionally the whole weighting for everything being done now would be gone to waste and the same with animations.

Almost all the content created over the past almost 10 years would be worthless.

What happens with old avatars, should we have both (new and old) in the engine? that just gives even more confusion and blows SLs system up even more, by keeping track of 2 different weighting systems, skeletons and all other internally handled procedures.

I know it always sounds easy, but you have a bunch of creators having spend time and money into their creations, wanting a perfect solution so not everything they ever made goes to waste.
And you have existing users, who don't care for mesh or new models and just want to be able to use everything they ever payed for.

Not an easy decision and surely nothing that could be switched within the blink of an eye. There have to be made a lot 'stable' solutions and decisions most users would be okay with at the end, before replacing one of the 'core' elements.


=)

[Rahkis Andel]

Haha, well, I won't repeat myself trying to retort that top part -- Let's just say that I'd be more than happy to see a new avatar be released. But I imagine LL will still be getting complaints no matter how good it is.

As for the whiny bit -- What I hear is that the avatar mesh as it is is deeply bound to the system.  The details are not for the likes of us, I suppose. As a programmer, I can understand what the problems could potentially be but I'm just guessing. It sounds like a whole mess of poor documentation and lack of foresight.

But hey, SL is an extraordinarily complicated project to jump into for a company so I'm pretty impressed that it's gone this long and not totally broken down.

Edit: Oh, yes Codewarrior reminded me that also there is the problem with all of the pre-existing content that a new avatar replacement would break. it's pretty unfortunate.

[WhiteRabbit0]

The deformer itself allows for the creation of Avatar 2.0, 3.0.. 5.0  whatever.  Because the deformer operates based off of the approximate shape of the default avatar, you can create a new avatar with it and simply include clothing layers with appliers, very similar to the lola tango setup.

[Coby Foden]

Code and Rahkis, thank you for the comprehensive explanations.  :matte-motes-smile:

I'm eager reader of different ideas, opinions and modelling techniques.  I try to suck the good things and excellent ideas like a sponge.  So I might eventually learn something new - even though sometimes the sponge leaks and does not keep all the info delivered.

 

Epilogue...

I wonder when we will get new beautiful avatar?
 

 

Beautiful, huh?  This should please everybody!
(and could you please stop whining?)
 

 

Erm... if you say so...
   ...... 

 

 

[Codewarrior Congrejo]

yikes ! 

what on earth.. no.. ...or ? i don't know.. i am confused.. can't stop looking.. i don't want to look like this.. or maybe.. I DON'T KNOW anymore.. that nose!!!...  *faints*

[Coby Foden]

---

Codewarrior Congrejo wrote:

.. that nose!!!...  *faints*

---

            

[Rahkis Andel]

---

WhiteRabbit0 wrote:

The deformer itself allows for the creation of Avatar 2.0, 3.0.. 5.0  whatever.  Because the deformer operates based off of the approximate shape of the default avatar, you can create a new avatar with it and simply include clothing layers with appliers, very similar to the lola tango setup.

---

Ah, I see. I honestly have no idea how it works, but that makes sense. I suppose after that, they just need to figure out how to get us a modifiable skeleton and we'll be golden.

 

What's 5 more years?

[Kitsune Shan]

Its kinda sad that most people here are more concerned about how the avatar looks on the DEMONSTRATION video rather than the feature itself. And this is why we cant move forward...

Personally, I want to see this project released as soon as possible. But I also noticed how some people takes so much time to calculate the deformations. Yes, it depends of the topology, but creators will still use as much as high poly they can. So doesnt matter how much effort you put on your clothes if near you will rez 2 avatars wearing 5 complete ZBrush projects without decimation. So I kinda understand if LL doesnt release the deformer due performance. The problem is that they dont even said if is due performance or something else and the hability for users to read LL minds isnt working as usual.

So, if this comes out as it is, we will surely see a lot of people complaining about the performance and time that takes some clothes to load and deform properly. I am even sure that for some people it wont deform at all.

I already suggested easy and simple ways to deal with this problem and make the project efficient. If someone is interested on see my last suggestion wich could fix forever the performance problem, just need to visit the Jira since I am kinda tired and bored of writing again and again while no one seems even to understand the benefits of it.

 

[Rahkis Andel]

---

Coby Foden wrote:

Code and Rahkis, thank you for the comprehensive explanations.  :matte-motes-smile:

I'm eager reader of different ideas, opinions and modelling techniques.  I try to suck the good things and excellent ideas like a sponge.  So I might eventually learn something new - even though sometimes the sponge leaks and does not keep all the info delivered.

 

Epilogue...

I wonder when we will get new beautiful avatar?

 

 

B

eautiful

, huh?  This should please everybody!

(and could you please stop whining?)

 

 

Erm... if you say so...

   

...

...

 

 

 

---

You'll be happy to know I'll be finished soon with your new beautiful avatar.

[Coby Foden]

---

Rahkis Andel wrote:

 

You'll be happy to know I'll be finished soon with your new beautiful avatar.

 

---

Amazing!  So much more detail than with present avatar.  :matte-motes-big-grin: 

[WhiteRabbit0]

I suggested this on the JIRA but I will repost it here for completionist purposes.

 

I do believe there should be a restriction on how many deformer enabled polygons a single avatar can wear specifically so calculation time is kept to a minimum, which does require some additional serverside coding from LL, which I eagerly await some, any, response from.

 

Honestly said Zbrush ports just about kill my poor PC when I have 3-5 AV's with them on my screen, I constantly de-render them when I see them, I wish there was restrictions on plain & rigged mesh too.

[Codewarrior Congrejo]

Amazing!  So much more detail than with present avatar

meeks...you've got to be kidding me.. *runs*

i am hereby addressing linden lab.. do not.. under no circumstances.. ever have Coby suggesting anything in regard of avatar 2.0 ! never-ever !

PS: all the fun aside - looks cute!  if you can port that look to the final version it even supports nicely that 'rough' wooden carved look of the original =)

[Codewarrior Congrejo]

The restriction might generally not be a bad idea to enforce more optimization onto the creators. I'm curious what comes out of your request.

Eeven though i still believe that 'knowledge' about how to create reasonable content is a big part of the issue, and why i am always so willing to give long answers on that matter.

I.e. with Z-Brush - it's not really the fact that the mesh comes from Z-Brush, it's rather an issue that the creator doesn't really know how and where it's use is in the creation pipeline. The content i am seeing is created in all sorts of 3D software and the outcome (addressing only the 'bad models' here) is mostly the same: way too heavy.

No idea what retopology is, or why it would be needed to make a lower version of a model and how to achieve this. And of course no idea about what  reasonable values are. 

Maybe the uploader should slap one into the face with even more complicated terms in regards of rendercost, estimated drawcalls and whatnot, or have numbers blinking in red.. so they wonder and start looking things up, and  to learn that there is 'more' to the whole concept of reasonable models and renderpipeline  then just how many prims it will take on a region. (just a funny side note -  not to be taken literally)


Just like i quoted one Dev yesterday, stating on the wiki for culling, that they somewhat did a bad job on teaching people about creating reasonable content, due to the belief that this should be something the creator shouldn't need to worry about and the software just deals with whatever is put into it. A philosophy that didn't really pay off at the end as he figured : )

And i have already said this in another thread, but i don't even think that normal maps and specular will relief us. I am kind of certain that people will just plaster those on top of the high polygon models.

But time will tell =)

[Alicia Sautereau]

If for the deformer to work as intended a new avatar has to be created: burn the deformer

Leave my avi alone