The application of muscular systems for the aims of CGI was ( and still is ) a priority mostly of the big film studios working on huge projects. But times are changing and the number of CGI artists and developers that are looking for new ways of building them ( musclular systems ) is increasing. It's a complex task as a whole and there hasn't been found yet a good decission for mass use that would be giving good results.

My opinion is that computer graphic and animation do not need muscular systems which give exact physical models of real muscles behaviour for they themselves are a visual illusion ( computer graphic and animation ).

I see the appropriate decission for mass use to have several basic qualities:
- first and most important - muscles have to preserve the body bulk when folding, unfolding and twisting limbs and torso. This is mostly important for shoulders/armpits and pelvis/thighs.
- reliability - to firmly "hold" the binded geometry;
- relatively simple realisation and easy at work - to use if possible the available instruments for muscle creating;
- flexible structure permitting easy coping with arising problems and quick adapting to different body shapes ( biped, quadruped, invertebrates, etc. );
- automated process of creating and editing that enables building in of muscular systems in numerous virtual characters during a relatively short period of time.

This muscular system offers three basic ways of using it and dynamic skin and/or dynamic muscles could be added.

This is the basic working method when building even working muscular system with high quality.
The idea here is to use muscles as geometry based smooth skin influences. This enables the bind geometry to "act" firmly and to be controlled precisely by them. In other words, how mutch skin surface will be attached to every particular muscle and how strong will be the muscle's affection.

After the skin weights have been adjusted, there are usually no unpleasant surprises on some of the stages that are to follow.

It seems many artists underestimate the strength of the geometry based smooth skin influences, but according to me it's the main precondition for building realistic and safely working character rigs. This method secures a solid base to which almost without any problems could be added extra deformers, extra controllers, dynamics, etc.

In fact as everywhere else there are also some refinements. After a long period of wandering I seem to have found the appropriate working formula:
- it is not necessary to build exact physical muscle models, it would be enopugh to create only the contacting surface of the muscle with the skin. This reduces the number of muscular CVs/vertices and thus the smooth skin cluster's performance is improved.

The above said are only references of course - it is quite probable that sometimes the muscles influence could be greater than usual but let it be mostly in cases when bodies with massive muscles and/or subcutaneous fat are placed in extreme postures.
Do not hesitate to apply really great deformations to the muscles, no matter that they do not look well at time of animation, important is only the resultant influence on the binded geometry.

The negative side of this method is that a really precise weights painting has to be used.

With the second approach the muscles are used as NURBS geometry based sculpt deformers.
My opinion is that this method is not reliable enough for serious projects, I myself have never used it in my work and I don't think that would ever count on it totally. But it might be a good decission for simpler muscular systems that aren't going into extreme postures and has to be built in short terms.

The method offers two basic advantages:
- sliding effect of the skin along the muscles;
- the prolonged period of skin weights painting can be avoid which as a whole is a serious advantage.

Also there are few disadvantages, arising mostly from the requirements of Maya for the geometry used as sculpt deformer:
- the building of much more precise muscles is demanded. Form, deformation and position has to be to a great extent adequate to the real ones. If not when in some extreme postures their reaction won't be the right one and geometry would "slip away" from the sculpt deformers;
- the starting and ending cross-sections have to be with zero size ( sx = 0, sy = 0, sz = 0 ), which means that each muscle will have 2 extra cross-sections, which increases the number of manipulations for "hooking" them to the skeleton. And besides each muscle used as a sculpt deformer must have a convex shape;
- for all sculpt deformers will be needed precise adjustment of maximumDisplacement / dropoffDistance attributes. In most cases the safe way is to set them to 0. The main reason for all these advices is the fastidious nature of the sculpt deformer. Why that you must be very carefull with the muscle shapes and the sculpt deformer's attributes. Also, another reason is, if muscles ( sculpt deformers ) + sculpted geometry are far from the world origin ( for example tx => 2000, ty => 2000, tz => 2000 ),
- the direction of the surface normals of each muscle's envelope has to be taken into consideration for this would be the direction into which it is going to displace the geometry.

The pictures that are to follow will enlighten these requirements and restrictions:

The third method is a combination of the above two, e.g. muscles are used as geometry based smooth skin influences and as NURBS based sculpt deformers.

The advantage here is that the geometry is held tight by the muscles used as smooth skin influences ( as i mentioned above: controls how mutch skin surface will be attached to every particular muscle and how strong will be the muscle's affection ) and using them at the same time as sculpt deformers enables preserving the body bulk at extreme postures, creating in addition sliding effect of the skin along the muscles under it. Of course it is not obligatory that all muscles should be smooth skin influences and sculpt deformers. By means of "test - mistake" the "recipe" for each one of them could be found.

For simulating ( and controling ) of the above mentioned sliding effect it is necessary for the muscles to have a greater influence as sculpt deformers than as smooth skin influences.

I myself apply this method to the areas of the armpits/arms and pelvis/thighs but only if the limbs are going to suffer really great twisting.

This approach is rather labour-consuming compared to the above two, but the results are really good.

It has been a preffered topic lately, but in my opinion using dynamic rigs in small and medium size projects is more like science-fiction. It's a fact, that most artists don't see how deep and serious a task it is and what a great amount of difficulties has to be overcame before reaching any kind of results.

My only advise is - use dynamics only in extreme cases where quality is first priority.

For recreation of dynamic interaction of skin and muscles together with their specific dynamic characteristics like flexibility and inertia, is applied soft body or cloth simulation onto the binded geometry. This brings a sagnificant betterment of visual quality but the process of adjusting the dynamic parameters is long and difficult.

Soft body - the idea here is that the combination of soft body and smooth skin cluster is working stable. As a rule, every virtual actor has 3 geometry layers ( animation proxy, binded geometry and render geometry ). It is a good idea soft body to be applied onto the binded geometry and not onto the final render geometry - this speeds up considerably the dynamic simulation and at the same time preserves the same quality of results. Of course, the muscles will be used as collision objects for the soft body's particles. After that, everything is PP attributes painting/tweaking and collisions adjustment.

Cloth - we know that the dynamic model of the real skin is much like this of cloth materials. That's why it is suitable to use cloth simulators for recreating dynamic skin. Maya's native cloth solver and Syflex Cloth Simulator offer reliable means for that purpose. Each one of them enables painting properties straight on the surface of the cloth object ( the binded geometry ). It gives a considerable control on the final result. The muscles, again will be used as collision objects.
As a whole Maya's native cloth solver is more precise and yet more slow and difficult to manage. Syflex Cloth Simulator is much faster and easy to use.

Here are some basic problems connected with the application of dynamic skin:
- in many cases will be needed very precise recreation of muscle and bone structure;
- replaying and caching of dynamic simulation has to be carried out for any scene containing virtual actors with dynamic skin;
- when making quick movements of the virtual characters supersampling is required e.g. computing of dynamics at subframe level or else the dynamic skin would react incorrectly. This considerably slows down the replay of the simulation and costs also a much greater amount of cached data and at the same time there's no guarantee that it would help with some really extreme movements;
- no matter how high the quality of dynamic skin setup is, almost always is needed extra secondary control at places where dynamic simulation has produced bad deformations. It is done by baking of dynamic simulation on geometry shape level, after which again are used smooth/rigid skin clusters or other deformers for applying corrections.

This is the preffered by me method for recreating dynamic interaction between skin and muscles - it's quick, easy to realise and control is secure.
The idea here is to use muscles as smooth skin influences and sculpt deformers and to apply soft body to their intermediate cross-sections. The main part here belongs to soft body's massPP and goalPP. In some cases, if a quick transformation of muscles is needed into different extreme postures, springs are used to compensate the unwanted deformations.

Let construction of muscular system be not a goal in itself, generally speaking they are not needed. Use them only in situations where are really needed and mainly in areas where only binding to joints/transforms is insufficient. Such areas are usually shoulders/armpits and pelvis/thighs.
Do not make things complicated, keep to the formula: less advanced technologies = less problems = better results in a short period of time. For me at least it really works.

Muscles won't help you finish your work quickly. They will help you do it better, but only if you devote to it sufficient time and great effort. And let's not forget the "hard" testing ( and I really have in mind HARD TESTING ).

Don't be quick at setting on fancy third party deformers before using up all instruments from "stock-in-trade". In most cases the external plug-ins bring more trouble than help - especially when there is a big render farm.

Basic concepts behind the muscle rigs

Some advanced tips and tricks ( still on the work table for now )

There are two main differences bethween the "Public Edition" and the original toolset:
- in the original version, the computations of the muscle components are accellerated with C++ code;
- a set of the powerfull ( but very pipeline specific ) functions, regarding creation and manipulation of muscles in batch mode, are not included in the public version.