This week I finished cobbling together my study of Maya Muscle versus the much simpler, more commonly used LBS skinning method. I ran into a number of difficulties with MM that made me understand why it’s not a commonly used skinning method, and did an animation test with video reference.
This term term and this term project really didn’t go as planned, and this study suffers from it: I didn’t get to spend as much time as I would have liked on this project, both on Maya Muscle and on the LBS rig, which I ended up omitting blend shapes from due to time constraints.
Here are the final video results side by side:
Notably, the LBS rig suffers from not having an elbow locator, which I only realized after I had reached the animation stage with no time left to add it in.
The past two weeks had been focused entirely on trying to troubleshoot MM, which was proving a finnicky tool to work with for various reasons I’ll elaborate on in my comparison between the two methods. Here’s my subjective breakdown of both deformation methods, focusing more on Maya Muscle, using the following criteria:
1. Reach Space & Range of Motion
2. Accuracy of Deformations
3. Ease of Implementation
REACH SPACE & RANGE OF MOTION
Both MM and LBS run into difficulties when the elbow is bent, albeit in different ways. LBS, as expected, runs into clipping issues. MM has elbow difficulties as well, where muscles that connect up past the elbow (specifically the ridge and flexor muscles in the forearm) do not actually bend with the elbow, but instead stick out past it. I attempted to cheese the issue by attaching the forearm muscles to the region around the elbow joint, but that resulted in the forearm muscles looking too small and deflated. MM deformed better than plain LBS did (though LBS would have benefit from blend shapes) in the shoulder area when the arm was raised, but the shoulder also ended up looking deflated when the arm was lowered.
ACCURACY OF DEFORMATIONS
LBS, as expected and discovered earlier, has difficulties around the bent elbow region.
Unfortunately, I can’t exactly evaluate MM on its skinning deformations, because when I attempted to bind the muscle to the model’s mesh, the mesh deformed in a very odd and destructive way:
This was something the tutorials I was referencing had not encountered or covered, and I wasn’t able to successfully troubleshoot it. Because of this, I’ll be evaluating the accuracy of the muscle deformations to the best of my ability instead of the accuracy of the skin deformations that would have been on top of them.
While creating muscle groups did help make a more realistic arm shape at T-pose, with a proper shoulder muscle that would accurately make the tops of the shoulders rounded and full, and some deformation relationships were accurate to the muscles (i.e. bending the elbow would squash the bicep while extending the elbow would stretch it), the degree to which the muscles deformed and the shapes they made while deforming tended to be very balloon-like.
The overall result is semi-accurate muscle-skeleton relationships, but sometimes very questionably accurate shapes. MM also notably featured some extremely odd jiggle physics that seemed aggressive for being default settings.
EASE OF IMPLEMENTATION
Here’s where MM really fails compared to LBS. LBS is known for its ease of implementation and straightforwardness; on the other hand, MM is a difficult tool to learn and use. The muscle system is complex, but lacks any significant documentation or tutorials. Throughout my attempts at learning MM, I found two YouTube tutorials, both from several years ago, as well as two sets of documentation from Maya 2010 and 2011. This means troubleshooting any errors I came across was extremely difficult or unsuccessful, and since nobody really uses MM, there’s not a solid set of best practices to work off of.
Maya also crashed numerous times while working with MM, especially when editing the shapes of the muscles themselves from a weird balloon shape into the actual shapes of the muscles. All things considered, I now understand why Maya Muscle’s results are not worth the trouble of learning how to use the system and implementing it; it’s heavy on the program, prone to technical complications, and unsupported both by Maya and by the Maya community.