Monday, May 16, 2011

how to make a face like this

i had the notion that the facial animation class was a tough, second-year kind of course.  not really sure why, and when i looked around the room one day in the latter half of the semester, i noticed that everyone was a first year like me, with one exception.  and there were also three underclassmen, too.

am i in love with animation and learning to sync speech?  no, not really.
was the class pretty interesting?  yeah, i'd say so.

part of what made it so cool was that my teacher was a pioneer in computer graphics and facial animation and was the guy who made the first computer animated face, a handful of years before i was born.  so it was fun to  just ask him questions and listen to the ensuing stories.

the class was set up for us to make two faces, which was a good method, since we'd discover what was wrong with our first design and could improve it for the second one.  my initial plan was to model chloe o'brien from 24, for not particular reason than i thought it'd be fun.
i found some nice reference images and started modeling my favorite computer analyst.

it was probably around 11:30 the night before this was due when i joked that she kind of looked like lord voldemort.
then i realized that, if i chose to make the dark lord, i wouldn't have to both with modeling a nose.
and at this point, chloe o'brien began to evolve into tom riddle.

you don't see you-know-who making a quizical expression often, but i think i did a decent job with him.
he's very blocky because that shows the geometric design of him, which is important for understanding if he'll be able to move and talk correctly.  in computer graphics, the geometric design is kind of like the muscle structure: you want things in the right place according to what it's going to do.
press one button in the computer and he comes out smoothed:

he had some problems, but i was able to make him say "happy" with moderate success.
(one of my friends make a really scary zombie clown and it didn't occur to him until too late that, without lips, it would be very hard for him to talk.  admirably, though, he did his best with that and it looked pretty good!)

at spring break we started into our second face, which would become our final project: animating full speech. i tossed around some ideas and decided to do "destro", the villain with the metal face from g.i. joe.
trying to find references for a fictional comic/cartoon character is kind of tricky when you want to make him in three physical and realistic dimensions.

this was kind of helpful, but hardly ideal references from which to sculpt.  eventually, i found this, a replica mask from some comic nerd website.

incidentally, these masks were never actually produced, due to low pre-order numbers.  still, i thought it looked seriously awesome; strong, powerful, imposing, definitely a mask but also very human.
now, our first model was due the day after we got back from spring break, and since i spent the whole time out snowboarding in colorado (never did get around to posting about that, did i?), this was my initial model:

mask: awesome.
this thing: goofy and dumb-looking.
but it was a nice exercise not only in matching to a model but also looking and see what was right and what needed to be improved upon: much bigger lips, a matching broader chin, shift the jawline, and, well, the list went on.
eventually, i got to this:

i would continue to make adjustments until the last possible moment, but he's certainly looking a lot more bad-awesome.  i adjusted the "collar" at the base of the mask and added a seam around the longitude of his head, where the mask would split into halves, just to add some realism while ignoring the question of how a metal mask deforms for speech or any sort of comfort issues.

i really would have liked to give him some cool textures for his face, some scratches on the metal, some spots duller and some spots shinier, etc.  but like i said before, textures are not my thing and we weren't too heavily graded on that, anyway.  so i just cranked the shininess levels and called it good (more or less....)

now, there are several different ways to do just about anything in computer animation, and we used a few methods in animating our faces.  one of the main ways is what's called "blend shapes."  basically, you make a lot of copies of your face and then sculpt each one into different expressions and poses.  this ranges from "right mouth corner up" to "left eyebrow down" and anything else you might want, as well as different speech postures: "oo", "ah", b/p/m" (all the same lip positioning) and so one.
soon, it looks like this:

then you press a few buttons in the computer and you magically have a row of buttons that controls those expressions on your face.
after that, you move through your vocal track and start matching up the mouth positions with the spoken syllables.  once that's done(ish), start working on the upper half of the face, adding eye blinks, brow movements, and whatever else to add expression.  one key thing is that the eyes should almost always be moving; motionless eye pretty much equals deadness.

since all of our faces will be shown together, we tried to have a common thread linking what they were saying.  in past years, they've answered the question, "what happened to your body?"
in maintaining the acknowledgement that we were working with floating heads, we asked "what were your last words?"
while i couldn't think of anything particularly creative, i decided to play against the stereotype and have my big tough guy's last words be him begging for his life.  meh.

a few weeks before the end of the semester i took a week and really focused, really pushed myself, and was super dang happy (and surprised!) when my teacher only a had a few--and very specific--suggestions for me the next day in class.  i added some lighting and a backdrop and also put a few objects off-screen to be reflected in his shiny metal face to make it more interesting to look at.  then, for funsies, i animated the camera to move like a documentary camera, adjusting the framing as he was talking (tried to do some focus work, but that was more hassle than i felt like dealing with.)  a few more touches and he was done.


and i finished without having to do any terrible all-nighters.
which was nice, because i inexplicably needed a lot of time to work on my compositing final....


Sariah said...

I wish there was sound w/that amazing video! But even without it, awesome!

The Former 786 said...

Yeah, what was he saying? To me, it looked something like, "Mommy, don't hit me. Whatever you do, I don't like the robots."

Which is pretty creepy, if I do say so myself.

--jeff * said...

of the three version formats that i have with me, one doesn't play at all, one doesn't have sound, and the third works but is too big to post on blogger.

i'll work on that when i get back to the lab....

--jeff * said...

joel, you're actually pretty close on the lip reading, by the way (although it's not actually that disturbing.)