I can move! I can walk! I can talk!
This week it's all about bringing things to life through animation, with musings on the potential crossovers with generative AI.
Over the last nine weeks of my Animation Basics class I’ve done a veritable collection of “Hello World” animation assignments, such as:
Three bouncing ball exercises
Two “tail” exercises (many tears were shed)
One pendulum exercise
All of which have culminated in vanilla and character walk cycles. We’re still too green to use a full body rig without hurting ourselves, and so we work with Ballie: a literal ball with two legs and limited joint articulation.
Substack doesn’t yet let me post videosso I can’t give you a frame-by-frame play of what I crushed and what I bombed in this assignment, but I can say that the whole thing gave me a newfound appreciation for human mobility.
Here I am, sweating over my keyframes and trying to get a simple ball with two legs to take four steps from point A to point B, thinking about the angle of rotation in the hips in relation to the length of the gait and the timing of the foot placement and the bend of the knee, and yet, as a human, I don’t even notice what goes in to my own walking. What takes me 10+ hours to animate in Maya takes me less than two seconds to complete in real life.
Human bodies, y’all. Amazing.
That’s just like, your opinion, man
We’re headed into the land of rambling thoughts and unresearched opinions - tread carefully
So let’s extend a simple vanilla walk cycle to generative art, because lately I’ve found myself wondering “will generative art replace 3D character animation?” And while I think the short answer is “not any time soon”, I’ve been trying to articulate why exactly I feel this way.
But first indulge me as I get a bit pedantic: there’s a difference between movement and animation.
Movement is as it sounds: something is going from point A to point B. We’ve seen generative AI models do this - most notably Meta’s text-to-video generator. Generative AI for movement, while imperfect, is here. There’s nothing wrong with movement - it’s fantastic! But it isn’t animation.
Animation is emotion and feeling filtered through a character, the so-called Illusion of Life. The character doesn’t have to be human - a good animator can make you care for a simple line, shape, or flour sack.
We could argue that a vanilla walk cycle - so named because you could use it on any character and they wouldn’t stand out in a crowd - is movement and not animation. Fine. But what about character walks? Could a generative AI system learn the rules of, say, a Charlie Chaplin walk and then consistently apply it to a rig for an elephantine character that’s learning to walk on two legs?
3D character animation involves a good deal of control, and it’s not uncommon to be setting keys every 1—12 frames to control a character’s poses and how the character moves from one pose to another. For reference, a 90-minute movie contains 129,600 frames, so even if we’re only setting a key every 12 frames that’s still over 10,000 frames manually being set by human animators, - not to mention the curve adjustments between each frame needed to control the movement.
All of these things combined make animation challenging for humans - so I’m not entirely bullish on a generative AI model figuring it out in the near future. Heck, even when we use motion capture suits for animation it still requires a team of human animators to go in and polish things up for production.
Now I could imagine a system where you feed in a bunch of Pixar movies to generate a Pixar-style animation, but at best I think you’d get a couple seconds out of each run (enough for a shot), but you’d still have an incredibly laborious process with expensive compute (animators are, by comparison, hella cheap). Not to mention that if you were successful with this approach but weren’t employed by Pixar you’d be staring down a wall of attorneys and lawsuits within minutes of hitting publish.
This is something I’ll be spending some time digging in to, and am always looking for more information. If you’ve got any reading and research recommendations please drop ‘em below!
One more thing
Even if generative models get so good that they can replace every animator in existence, I can’t imagine giving it up. Because part of the joy in animation is the joy in creation, of coming together with others to tell a story that makes an audience feel something. Take, for example, In Your Arms by Kia Grannis, one of the most charming stop motion videos I’ve seen in a long time. The corresponding “Making Of” video encapsulates everything I love about the process of creation and collaboration, something I don’t think can be easily replicated or replaced.
Release the drafts
What is the point of Bran?
Olenna Tyrell ❤️🔥 (there’s a link there but it’s not showing in my drafts)
Season 7 is far better than I remembered. A+ for drastically improved cinematography and clever little cuts.
Absolutely zero chance Bronn could operate the ballista by himself. None.
I've made some poor relationship choices in my life, but none as bad as Jamie Lannister.
Dream job: animating dragons and zombie polar bears.
The Rains of Castamere is a real bop and The National did it best.
Kind of wild that all of Westeros was led by leaders without a Google Drive folder full of strategy documents.
That’s it for this week! Thank you for reading, I love you!
a few days after writing this I was able to get into the Substack video beta program - I’ll be sharing regular, non-email posts of some of my work over the next few weeks, including homework submissions and mentor feedback!
Love the distinction between movement and animation. Just finished a chapter on movement (deceptively titled “animation”) and 100% agree.
WRT the concern about generative AI replacing animators entirely, I don't think that will ever happen. AI-based generators and tools will be tools first and foremost. Animators won't have to spend as much time on learning the framework's specifics but can spend more time on doing the thing they want to do: animating. And the iterative process working in more natural language will speed up the process all around and encourage more innovation. AI-based software generators (e.g. open ai's codex) will help software engineers and the like focus more on the engineering aspect rather than on software bugs and other such things.