This afternoon, I came across the following clip from James Baxter's blog:
It's a clip from Enchanted (2007), showing Giselle getting out of the carriage in that extra pompous wedding dress, complete with heaps of ruffles. On viewing the completed animation in context, it may not look like much... that is until you realise that, 1) someone animated this entire sequence by hand, 2) that dress has heaps of ruffles on it that all need to be individually drawn/placed, as the entire shot was drawn by hand using traditional 2D animation techniques.
All of a sudden, the complexity of his shot strikes home. This is even true if you're a CG artist instead. That's because, even though you not need to draw and place all those ruffles by hand, this is still a rather tricky shot to manage. Anyone who has any experience with cloth sims know that they can be finicky to manage - even the slightest differences in poses, timing, and settings can be the difference between the sim "exploding" in dramatic fashion, doing exactly what you want, or stubbornly refusing to hit the poses you need (and with little way to art direct it to do so). And that's already quite bad when you've only got a relatively simple piece of cloth (i.e. like a silk body-hugging dress)... having something with that many ruffles on it just means you've now got a self-collisions nightmare on top of that, which now takes ages to compute. Oh, in case you've forgotten, there are some character <-> cloth interactions here too (i.e. hands/fingers and feet in particular), which also requires quite a bit of back and forth to make sure they remain in contact.
So, if you're a CG artist who's just spent ages trying to tame a cloth sim, you may think - "you know, it seems they had it easy back in the 2D days... if you want a certain pose, you just draw it yourself". Well, think again:
"To animate them, I just animated the edges first, not the vertical lines of the folds, just the wavy lines that form the edges of the ruffles. That way it was just like animating a springy wire- just a single wavy line for each ruffle. I did them one ruffle at a time, concentrating on keeping the design, making sure the big folds stayed big, the squarish folds stayed squarish, working straight ahead to get the overlap looking good,while referring to those major design keys as targets. I went through sections of the shot in waves, adding a new ruffle edge each time. Then, when the edges were done, I filled in the vertical lines of the folds. This was easier, as they just had to follow the animation of the edges that were already there.-- Source: http://jbaxteranimator.tumblr.com/post/115194646099/hi-i-was-wondering-if-you-had-any-insights-or
Uh huh. That's some animation required! Serious respect is due here indeed :)
While we're on the topic of complex animations, I must also compliment Hjalti on the complex character/prop interaction shots he's been working on for the Gooseberry pilot. Watching the timelapse was an insightful experience (Note: I've been waiting years to watch some of these timelapses of animators watching. Keep them coming ;) But the latest version of that shot when he's taking the rope off and putting on the timer is even more awesome. I really love the interplay between the characters in those moments where Victor is kindof dragging Franck towards him, an especially Franck's expressions when the belt is being tightened! It's like that cute moment in Ratatouille when Remy passes the pot of Linguini-soup, and we see his face just briefly bloat up for a frame or so in disgust. So much fun!
Truth be told, I've got a few ideas brewing about how exactly we could develop tools to make these kinds of complex interaction scenes a lot easier to accomplish for the animator. I've been dying to have a serious crack at them for the past few months, but currently can't find the time required to put these into motion with all the other important projects on my slate right now. The gist of these is that there are a wealth of possibilities to explore where we start coupling interactive ways of just "telling" Blender what the poses should look like and/or move, getting some "physical awareness" about the things we're dealing with, harnessing the power of sims as a guide/aid for interactive tools, and doing away with complex rig hierarchies with a bazillion switches. I'm being purposefully vague here, though I suspect that with what I've said here, those in the know about all this stuff actually have a bit too much information about how to go about this already ;) (If you do, get in touch!)