Why isn't it possible for software to automatically make well-fitting mesh clothing?

[Jennifer Boyle]

I have no knowledge of the actual process of making mesh clothing, but I do have some understanding of computing and data-handling principles.  I have developed a love-hate relationship with mesh clothing.  I love it because some of it is absolutely gorgeous, particularly skirts, dresses, and sweaters, and I hate it because so much of it fits so poorly.  I understand how hard it is to make clothing to fit the huge variety of shapes.  Then I started thinking that, in principle, it should be possible to fit each shape exactly.

The viewer, presumably, knows the exact shape of each avatar that it sees up close, and certainly knows the exact shape that the avatar of the account it's running on has at the time.  Therefore, a viewer, or a specialized program connecting to the server in place of a viewer should be able to capture that information and use it to alter a mesh garment to exactly fit the shape.  I realize that there is some complex analytic geometry involved in this, but it appears that nothing that is ground-breaking would be required.  It would really just be calculating the three-dimensional location of a large number of points in one system and making corresponding three-dimensional points in another system have the same, or, perhaps, slightly offset, locations.  I could visualize either a product that consumers could use to alter purchased clothing or one that creators could use to make each customer's clothing fit precisely.  Either way, it would solve a big problem and make my second life better.

After thinking about the concept above, I couldn't help but wonder why we cannot have an in-world drag-and-drop editor for mesh clothing (and for prims, too).  Again, it would take some complex programming, but not anything ground-breaking.  I imagine that drag-and-drop editors for two-dimensional representations of three-dimensional objects must exist.  If that's so, and it can be done in other contexts, why not in SL?

I just cannot understand why these things, which would be so useful, are not possible and why no one has done it.

ETA:  Thinking soime more about these concepts, I wondered why there cannot be a drag-and-drop editor for system clothing.

[Dilbert Dilweg]

We are all waiting on the mesh deformer

http://www.qarl.com/qLab/?p=80

[hibit Spad]

Part of your answer is in how poorly made the avatar mesh is.   When LL made the avatar mesh back in  2004 or there abouts, there was no thought given to what they might eventually want to do to it or with it in the future.  The result was something that worked at that time but now limits creativity.  Take a look at how the vertices of teh mesh move as you resize the avatar and also look at what happens to the avatar as you move it through some yoga like poses.  The legs will deform in very strange ways.

 

 

[Gaia Clary]

Certainly the Avatar has issues with its mesh, its unwrapping and its weighting. While in principle all of that could be fixed, all fixing would certainly have influence on existing shapes and clothes. So i can understand that the Avatar is not touched by Linden Labs.

However it looks pretty much like it just needs a few fixes on the skeleton side (and on the mesh importer of course) to get much more freedom even with the existing Avatar and skeleton and without breaking "old" content.

Here is a blog article i currently write together with Magus Freston, who has created the major part of Avastar.

    http://blog.machinimatrix.org/avastar/the-second-life-skeleton/

That might give some insight on why things are so complicated and hopefully it will also motivate others (intensive staring towards Linden Labs development department) to add their knowledge so we eventualy get a complete picture about what the Second Life Skeleton is all about

But beware, right now the article is only half done and its a bit mind boggling in some places...

nevertheless enjoy

[Nalates Urriah]

I'm not sure we understand the question your asking. You seem to think software could be made to automatically make clothes. That is possible. It has never been done because no one wants it. Blender is for designers to create clothes that have never been thought of before. Machine design tends to standardize.

You may be asking why clothes just doesn't fit? If so that is simple. When SL was created they made the shirt, pants, and other things we WEAR, which fit perfectly. The mesh clothes that we ATTACH were not planned for and we have fit problems with them. You will hear many say the Lab did not complete the implementation of mesh. While that is a rather ignorant simplification in some ways that is an accurate explanation and could answer your question.

The more complex and real answer is in the way 3D models are built. Just as a word processing program manipulates letters so too a 3D modeling programs manipulates vertices (points represented by numeric attributes). AS a word processor knows about words, sentence structure, and paragraphs so too does a 3D modeling program. It knows about circles, cubes, materials, and such.

Word processoers never tell you when to break your writing into paragraphs. They can't because they can't deal with the 'subject' of your writing. They cannot tell when your idea completes and a new idea or thought is being expressed. 3D modeling programs do not know about cloth. You can tell them to treat a group of vertices as cloth. You can tell a modeling program that an avatar body, which is another group of vertices, is an object the cloth collides with and is not to penetrate. It is possible to so designate those things in Blender or 3D Max. But, those features have not been built into Second Life. We are just now adding them. That is what the deformer is about.

Even the deformer doesn't recognize clothes. It sees a group of vertices. It knows that it should move those vertices in special ways according to certain rules. So, while dresses can follow the movement of our legs, they cannot swish and swirl like cloth can. Flexies are a way to fake cloth movment.

While you may see cloth in other games, there are no games I know of where user created content can have real cloth simulation. Even in Blender with CUDA and the GPU helping rendering cloth behavor brings my quad core CPU and 560 GPU to their knees.

Said in a more technical way... SL is being retro-fitted to render specially designated groups of vertices attached to other specially designated groups of vertices according to distances and relationships established by 3rd party designers that are to be added to settings estabished by users. No where in this equation does the idea of clothes and FIT even start to appear.

So, while it can be done, the actual task is far more complex than most people think simply because computers have yet to be taought what clothes are... 

 

[Jennifer Boyle]

I am not asking about software designing clothing; I am asking about having software fit clothing.

Here is a RL example of something analogous.  A few years ago, some clothing manufacturers tried something called mass customization that never took off.  They had machines in their stores that could scan a person and record her exact size and shape.  They kept the information in a database.  The customer could go online and order a clothing item, and, through an automated manufacturing process, they would make the item to exactly fit the customer.  Since it was all automated, it was low-cost.  The machines had nothing to do with design, but they made it fit.

Now, about SL and what started me thinking about this concept.  I find mesh clothing to be both lovely and exasperating.  The two worst problems I have with fitting it are the waists of pants and low-cut backs of tops.  I often have to wear sizes that are otherwise too small to avoid having a gap between the top of pants and my body.  Sometimes, I just can't get any size to work.  Unless I wear a small size, tops that have the top of the back going across my back have a gap between my body and my back.  Why can't software  know that the waist of a pair of pants or the top of a dress needs to be no further than a few millimeters from my body surface, and alter the shape of the mesh accordingly?  After all, my avatar shape is recorded somewhere as a bunch of three-dimensional coordinates, and the mesh shape is also recorded somewhere as a bunch of three-dimensional coordinates.  Even if the software needed some user input, i.e., what part to alter, or how close to the body surface to make it, the concept would still work.

To end on a lighter note, I've just discovered some rather hilarious effects when I wear mesh clothing and avatar physics together.

[Galathir Darkstone]

In all seriousness and with no malice or disrespect intended, your opening line highlights the reason you don't understand the "Why?" -> "I have no knowledge of the actual process of making mesh clothing"

There is no mechanic in place for the body shape deformations to be copied from the avatar to the rigged mesh. Hence the need for Qarl's deformer... or something similar (for any interested, you might want to have a look at this thread on SL Universe: http://www.sluniverse.com/php/vb/content-creation/73983-alternative-mesh-deformer.html ).

The short answer is that it is possible to have software automatically adjust the rigged mesh, but necessary elements aren't in place yet, and even when they are, there will be limitations in how well the rigged mesh will adjust depending on how different the rigged mesh is contructed compared to how the SL Avatar is constructed. This is a fundmental limitation when dealing with digital objects that have very finite vertices, vs the real life comparison, where human bodies and fabric have many, many, many more 'vertices.' Not to mention fabric in RL clothing can slip on a human body, while rigged mesh certainly cannot on an SL avatar... it really only stretches based on bone weighting.

[Jennifer Boyle]

Thanks!