I was asked recently about how I go about polishing. There’s no doubt about it, polishing is tough! I still struggle with it a lot, but I’ll try to lay down here some of what I’ve come across that has helped.
First off, I highly recommend Victor Navone’s Splinophilia tutorials on his blog: http://www.navone.org/HTML/Tutorials.htm
Victor was my first mentor and although I can’t work purely in splines like he does, learning how to clean and control them from him was really valuable.
When I’m blocking out a shot, I work in linear. I learned this from Keith Sintay who recommended it as a faster way of working. It helps with making sure that weird gimbal rotations aren’t occurring and causing problems later down the line, unlike stepped. I found that when you are posing, which axis you rotate in (x, y, or z) and the order in which you do it can really affect how one pose transitions to another. Working in linear, for me at least, helps with seeing the actual timing of the transitions between the poses. I find that with stepped mode I have a harder time judging how long a pose will hold. Also, if you work in linear and use copied pairs, then you can see exactly how long a pose will hold and how long it will transition to a new pose.
Once I have the major poses in and have keys that lock them off for the time that they hold on screen, then I go into breakdowns. I’ve found it takes a lot more thinking and time to get them right than what I thought when I had started animating. Basically, I’ll sit and think about how I want the transition from one pose to the next to look, concentrating on the main body mass and the head – Do I want it to slow out or slow in? What kind of arc do I want it to have? According to the mechanics of the move, what needs to move first and what will overlap or follow thru? Those questions will generally get me the pose that I’m looking for. Eric Goldberg does a detailed explanation of this in his Character Animation Crash Course. He explained that you use the breakdown to define the characteristics of the motion. Really worth a read/purchase if you don’t have it already.
When I have a good idea of the look of the pose I want, I scrub between the two keys and choose a key frame that has most of the pose in roughly the place I want, select all the controllers (use the gui if it works for your, or make a shelf button), then key it. Then I slide that key to the right frame for the timing I want. I take it and adjust the pose so that the arc between the previous key and the next key is better defined, the limbs are overlapping or leading how they should, the pose works in silhouette, etc, etc.
Then I just keep on working like that until the end of the shot or the section I’m working on. If you work this way, keep in mind that curve overshoots are your enemy. Also keep in mind that if you want an overshoot to put it in BEFORE the held pose. If you key it after the held pose, then you’ve changed the timing of the hold and it will throw stuff off.
I break things down fairly far – 2s or 3s depending on the shot. Occasionally I’ll let it go longer, but I usually end up keying closer anyway. Also, I do any IK/FK switch or something similar over one frame. Blending the two is a giant pain in my rump and I’d rather just struggle getting those two frames right than trying to deal with how the curves blend together.
Once I get to the point where everything is pretty much set the way I want, then I go into the graph editor. Drew’s recommendation that I select everything and hit Plateau tangents has helped me quite a bit, and it’s not a bad way to start splining. Other things I’ve done are to convert them all to Flat tangents, then use a Tighten Tangents tool similar to the freely available Comet tool, which is awesome. Both methods get things pretty close to where it needs to be.
In the graph editor, when I’m examining the curves I have to remember what I’m using it for. I use the graph editor as a visual guide to help get the fractional movements on the poses right, to smooth out the arcs better than I can by hand, and to lock down things that need locking (usually feet to the ground). I’ll take about a 15-20 frame chunk and examine how it looks with the graph editor open. There’s a button where you can frame the graph editor view to the timeline selection. It helps a ton. Also, if the curve looks flat, scale it in the up-down axis until you are positive it’s actually flat. I start with the main body controller and again, sit down and think about how I want it to transition from one key to the next, and how it passes through the breakdown. I consider how hard I want it to hit, or how soft, and how fast or slow I want it to leave its starting position. Then I adjust the curves. I try not to touch the keys themselves as much, but I’ll sometimes move a key to make the curve more of the shape I want it (going back to what Victor taught – knowing how the slow-in/slow-out transitions look on the curve itself and how the steepness of the curve affects how fast it’s transitioning at that point in time). Always have the shot camera visible when you’re moving a key. This will tell you if what you’re doing is a subtle adjustment that you’ll just feel, or if it is a huge movement that will throw everything off. This is also the point where I’ll break tangents and free their weights to pull them in and out to make crests and sharp hits like I would for a bouncing ball assignment.
Once I go through the main body controller, I work my way up the spine, mainly focusing on the drag and overlap. I try not to shift the curves for overlap until the very last minute, and only then if I can’t get it in the posing itself. After that, I go and track the eyes, the wrists, the ankles or knees, and the fingers on different passes. Those I mostly will go through in a straight-ahead method and key frame by frame to get the parts to do what I want. It’s helpful to keep in mind that if you’ve now set your keys to flat or something, then every time you create a new key, it will change the curve slightly in the graph editor. Flat is the worst offender and can really throw off the rest of the curve if there is a lot of frames between your keys.
For the feet and legs, I like to keep in mind a few things. On the ground, I like to use footroll and when it leaves I like to switch to X rotation, transitioning over one frame like an IK/FK switch. Recently, I’ve ignored the curves on the foot controller more and concentrated more on how the arc of the knee is moving. What I’ve noticed is that in a walk or a jump or in kicks, it’s more helpful to me to think of the knee as the driving force and the foot as overlapping. When I studied ju-jitsu, they told us to aim the knee at the target and then snap out the foot. I saw in reference videos that when people did jumps, they’re knees would point where they were going to land, and the foot would then follow into place before landing. I think conceptually it works better posing out the legs and it helps keep the knee pops to a minimum. Along the same lines, when you’re posing out the foot, the rotation of the foot and/or the foot roll will affect where the knee placement is, so if you adjust the rotation on the foot, you’ll want to take a look at how it affected the knee arc and adjust the foot translation accordingly.
I think that’s all I really have for you. At this point is where I playblast and just frame-by-frame things until it gets to how I want it, or close to how I think it should work.
This site helped me a lot in transitioning my thinking about overlap, follow thru, breakdown posing, and how it all related to Maya curves – Victor Navone had posted it in his blog: http://fliponline.blogspot.com/2007/04/quick-trick-simple-approach-to-overlap.html
Hopefully that’s of some use to you.