PDA

View Full Version : fEATURE FILM FACIAL ANIMATION


RenisanceX
07-03-2005, 02:04 AM
fEATURE FILM FACIAL ANIMATION PRATICES

Hi everyone. Im doing some facial animation and was wondering what is the typical proceduren used in facial animation for film more targeting the vfx practices

T4D
07-03-2005, 02:47 AM
these so many different ways it's amazing !!
you really need to do some google time dude

the Gollum in LOTR to my surprize it was all morphs targets in Maya
But not only did they make morphs for the main Smile, Sad, OO etc
they ended up making inbetween morphs to and from every other morphs !
talk about hundreds maybe thounsands of morphs :eek:

then you have the boned face rig where you move control object ( bones can be used ) on the face to get your poses see max guy uses this alot and i have a plugin for this method for LW , Also read it's used for Mocap face rigs as well.

then there's the bone muscle things ( high tech stuff ) model jaw & skull put muscle in between deform the skin around them,
Shrek was done this way, PDI really get into this stuff
I think even way back when they did it in ANTZ they used this method
tho i feel Gollum method looked better.

then there's all the way to control the facial animation
Control objects, sliders, Joysticks, or just straight Graph editor as some animator prefer.

if you interrest alot in this stuff Buy the book Stop Starring By Jason Osipa
it's THE BEST !!

ANGEL_OF_WAR
07-03-2005, 05:00 PM
I've heard really good things about DAZ Mimic for Lightwave (www.daz3d.com (http://www.daz3d.com)) but I can't give any personal examples myself.

RenisanceX
07-03-2005, 11:01 PM
Thanx im not to bothered about the app just methods. Has anyone have any information on how they did they hulk...i really loved how that characters face emoted. I searched but i found little info and if any it was based on his body's muscle system

ranhell
07-03-2005, 11:41 PM
:) Hi RenisanceX If you have light wave I would use morph targets and Bump maps tied together for more detailed results the reason for the bump maps is that you can squeeze a little more detail out of you’re model with out having to ad more geometry....

Now for the hulk if you have the DVD watch the ILM making of the hulk. They used allot of methods to get what they wanted. They used three motion capture cameras around the actors face to track the markers on his face to get those emotions they did the same method in the matrix

RenisanceX
07-03-2005, 11:56 PM
ranhell and how were these saved as morph targets? so essential they created morphs?

PauHana
07-04-2005, 02:59 AM
Actually, most of the "making of" features on DVDs are just fluff. My co-worker worked on The Hulk and a lot of the data they got was unusable or just looked bad. Which left them to hand key a lot of the shots. There's at least a couple other movies that people I know have told me that the methods they say was used was nothing more than a test that ended up not being useful.

T4D
07-04-2005, 03:34 AM
if you've only got LW you pretty much only have Endomorphs to do the job
I have done many face bones rigs in LW But I used them Over a simple endomorphs rig
so i can edit in Layout and not have to go back to Modeler each time i need alittle extra motion here or there.

the bone muscle thing Simply Can't be done in this version of LW.
SDK issues, Plugins Issues, layout speed issues, layout mesh editing issues. there a long list there, If you need it, XSI is cheap at the moment.


the Mocap bone/shockmonkey rig Have been done in LW But I presonall havn't never seen them in action, AND My only real experance with Mocop is with motion builder and didn't do face stuff so can't help there. as well ..

endo's are your main option, the next thing you need to work out is how you want to animate the face..

pooby
07-04-2005, 09:42 AM
I would Highly recommend getting Smartmorph. With it you can use an effectors to position (and animate) the influence of your morphs on the face.

Unfortunately, as much as I like XSI, you can't do this method effectively. so I'm still using Lightwave for facial stuff.

By using this, and bones for jaw movement, you can get phenominal control over the face with suprisingly few morphs.

If I get time, I'll post an example.

pooby
07-04-2005, 05:39 PM
Here's a simple example of smartmorph's abilities with one morph target.

http://www.movinghousefilms.com//uploads/misc/face.avi

RenisanceX
07-04-2005, 05:50 PM
poopby i think that would be easy to set up in xsi

Im not worried about software i have my choices Im a student.I can work is xsi max maya and lw so its doesn't matter more interested in techniques so keep them flowing

@ PauHana I do agree i also heard alot of things on the making of dvds are basically alot of fluff besides except for lotr extended can't remeber a movie that real does explain alot ...mabye matrix

twidup
07-04-2005, 06:18 PM
from an animation stand point, you can also rig the face with just joints...enveloping the skin to the joints, no fancy muscle joints, just regular joints. then you just simply translate/rotate them to teh desired position.

This give teh animators way more control than straight morphing....of course, with maya/xsi/max you could set up some set driven keys to control a couple of hidden joints, or morphs as well.

-todd

ThE_JacO
07-04-2005, 06:38 PM
the Gollum in LOTR to my surprize it was all morphs targets in Maya
But not only did they make morphs for the main Smile, Sad, OO etc
they ended up making inbetween morphs to and from every other morphs !
talk about hundreds maybe thounsands of morphs :eek:


hey Thomas, there's a slight misconception there.
Gollum used a technique that Bay called combination sculpting (which pre-dates Bay's work, but he agreeably brought it to a different level), which happens to be the same we used for two creatures when I was working on Brothers Grimm.

it's NOT based on full facial morphs, but it consists of modeling morphs that represent the main muscle masses of the face moving individually.
once you have these basic shapes nailed down, which are usually between 100 and 300 for a fully expressive creature, you start working on their combinations.

imagine having a muscle in the eyebrowse and, for that single muscular area, having a frown AND a lift, you then model the ideal combination of those two, instead of leaving the computer to interpolate it, and then rinse and repeat for several muscular areas that can be considered the main masses up to 3 or 4 way combinations.

this is what raises the count, from the couple hundreds basic ones, to a much higher number.

the face system then is presented to the animators with only the basic ones, and a simple polinomials based system will split the load and swap targets according to these, keeping animation and interpolation more human and controlled then what you'd get by simply averaging normalized shapes.

This is VERY different from having shapes and sliders for "smile" or "OOH".
if you want a smile with that system you will do it with sliders like "mouth corner crease Left" + "mouth corner crease Right" + "Upper Lip Curl" + "lower lip receed".
sometimes a few presets are made for animators to block out quickly (which will set and key a bunch of sliders in a basic expression or phonem), but ultimately the fine animation work will be done on controls for muscular entities, and NOT on things like a "smile" or a "sad".

Gollum had roughly 800 shapes by the end of the third movie, and the control panel featured nearly 200 controls (although most animators, some of them worked for us after LOTR, mostly used 60 or 80 of those and only used others for particular expressions or for the last pass of animation).

the advantage of combination sculpting is that it's presented to the animators exactly like it will be rendered, which for such a subtle thing as facial animation is a big plus, the downside though is that it's not very portable between creatures, and is hardly ever used to this extent for anything but hero characters.

pooby
07-04-2005, 07:02 PM
"poopby i think that would be easy to set up in xsi"

I put together a setup like it in XSI, but adding more than about 3 effectors causes such a slowdown that it becomes unworkable.

Megalodon
07-04-2005, 10:50 PM
And there is also TAFA - Timothy Albee Facial Animation. It's not cheap, but apparently it does an incredible job corrdinating facial animation.

http://ta-animation.macreitercreations.com/

Good Luck!

Vojislav+Milanovic
07-04-2005, 11:31 PM
TAFA did awesome job on Kaze Ghostwarrior movie. There more on in @ newtek profile on Timothy Albee http://www.newtek.com/lightwave/profiles/TimothyAlbee/index.php

T4D
07-05-2005, 01:40 AM
THANKS ThE_JacO for the info Really cool stuff :bounce:

I do freelance rigging in Lightwave ( adding XSI soon )
and do a afair bit of facial rigging I prefer Joystick control systems
But many animators don't
What did the gollum rig uses sliders or joysticks or something else ?

thinking setting up all that would be alot easier with joysticks
But WOW I still blow out thinking about the amount:eek: there's alot to deal with
the biggest Jotstick rig I've had only was 18 joysticks around 130 morphs
and I was thinking I was over doing itLOL :D

I put together a setup like it in XSI, but adding more than about 3 effectors causes such a slowdown that it becomes unworkable.-pooby

are you sure you set it up right ?
I have setup these in XSI and it all works extemely well on my system ?

ThE_JacO
07-05-2005, 01:52 AM
I do freelance rigging in Lightwave ( adding XSI soon )
and do a afair bit of facial rigging I prefer Joystick control systems
But many animators don't


joystick systems are basically banned from most hero characters in large pipes for 3 very simple reasons:
1) they introduce screen overhead and require view managment that could very well interfere with the cameras locked off from the animatics
2) they generate function curves for things like position or rotation which are not humanly readable.
every good animator I've worked with WANTS function curves and wants them on a normalized and consistant scale that he/she can understand and tweak in a graph editor.
3) joystick controls that drive hundreds of shapes are impossibly large, sliders can be well organized in a tabbed and grouped control panel.


What did the gollum rig uses sliders or joysticks or something else ?


mostly sliders to my knowledge


thinking setting up all that would be alot easier with joysticks


what is most used is not necessarily what is best and you could be right, but when you TD for many animators and they laze out from reading a 200 pages document for the rigs you go for the KISS rule.


But WOW I still blow out thinking about the amount:eek: there's alot to deal with
the biggest Jotstick rig I've had only was 18 joysticks around 130 morphs
and I was thinking I was over doing itLOL :D


can't disclose what, and I haven't worked on it, but I can tell you that a recent movie featured a character with well over 1500 shapes, several wrappers and deformers, 15 bones, secondary animation, and the odd scripted expressions/scripted operators to manage some things ;)
this is just for the facial system.

the more we progress the more "overdoing" fades off the vocabulary.

RenisanceX
07-05-2005, 02:57 AM
hmm thanx guys

Jaco is it kong :) lol

Any who I finished reading up stop stairng which was a good read joystick mead seems nice but control is and if for me..Maybe i need to do more research on it

Now i also did some research and Adam sales techniques called the muscle method which is based on enveloped curves situated all over the face to mimic major muscle groups Seems fine but i don't know or see what problems it could develop. It seems nice as int he face that you can also use it to create keyshapes. But not have experience in this area I am wondering about the draw backs

Also did the hulk use a similar system to gollums?

T4D
07-05-2005, 02:59 AM
Again THANKS SO MUCH :thumbsup:

Most of my work would be classed as "Toon"
so Joysticks have some advanages in the workflow
but don't completely understand the slider reason.

1/ and 3/
in XSI you can setup joysticks a tab menu driven view ( I've havn't done it but have seen it But not work with it LOL :D ) looked just like a slider panel view cool.



thinking curves in the Graph editor -
the slider/curve the animator would be controling would be a expression or x channel from the process of mixing the morphs would it ?
sort of like a joystirck is ?

But I do agree and feel the main reason is the animator would want complete total over what he has to work with. sometimes with joysticks you do play/fight alittle with joystick trying alittle too hard to get that inbetween bit " just over there but with alittle from back there sort of thing ":curious: in that case i jump in and edit a morphs or move a face bone But you can't do that at that level so Slider rule it.

anyway Top stuff, loved the info :love:

ThE_JacO
07-05-2005, 03:29 AM
Most of my work would be classed as "Toon"
so Joysticks have some advanages in the workflow
but don't completely understand the slider reason.


as I said there really isn't a right or wrong way of doing it, this is even more so if you animate your own stuff, which is where what is more comfortable for you is what is right.


1/ and 3/
in XSI you can setup joysticks a tab menu driven view ( I've havn't done it but have seen it But not work with it LOL :D ) looked just like a slider panel view cool.


what you've seen is, I believe, Graham's JS pages experiment.
it's all fine and dandy on a theoretical level, but if you have 200 controls it's not very practical rendering an interface into HTML contents and constantly reading them and monitoring them :)

speaking of overhead...
imagine 200 base shapes, divided in 4+4 groups, driving another 600 combinative options through polinomials that range between 2 and 5 solvable elements.
as impressive as it sounds, the truth is that I usually manage to reduce the mahts for such a system to less then 1500 lines of code (mostly I/O wrapping work) and only additions and subtractions in it.
the overhead nears zero, and the system is about as fast as it would be if you were to animate the shapes directly, as the sliders drive the base shapes directly, and the combinative system only kicks in in a hidden layer between sliders and choice of targets.

however, if these 200 shapes were to be driven by 20 10 way joysticks, to find the driving shapes and average them from the positions of the joysticks I would already have to monitor and crunch numbers for those that would be tenfold the load before I even get to the point where I figure out the retargeting, and then the retargeting (for ease of debugging and structuring) would have to analyze the joysticks again instead of just performing additions and subtractions on normalized values.

last but not least, designing and mantaining a system where EVERY value ranges between 0 and 1, and has a single axis progression, is enormously easier then managing one where I also have to translate positions and rotations into such values.

one less thing that can get broken.


thinking curves in the Graph editor -
the slider/curve the animator would be controling would be a expression or x channel from the process of mixing the morphs would it ?
sort of like a joystirck is ?


let's take a very simple example, I could have one 8ways joystick for the left mouth corner and one for the right, each would give me creasing up and down, extension to the left or right and contraction in the opposite direction with relative vertical offset.

these 8 possible base shapes would be extrapolated from the position of the stick, but what you animate is actually the stick, which is presented to you with 2 translation Fcurves (let's not even consider an eventual depth and pretend we have that as a wrapper) that pilot 8 shapes...
once I pose my mouth in a slightly creased up and compressed smile, how do I know EXACTLY how much of those 2 shapes I have in? and what if I want a crease up smile like that but ALSO want to add a bit of inward compression to beesting the lips?

this last case would be impossible, because the stick in an upper outward position logically prevents me from also having it contributing with the inward compression located in an opposite side of the stickpanel.

also, what combination of these values is exactly generating what shapes? and how, if needs be, can I exagerate this shape beyond a normalized 1 without having to let the stick loose and therefore creating a rigging exception?

with sliders that all range between -1 and +1 or 0 and 1 (and can eventually go in overdrive typing the value in), I can give animators control on ALL elements that will all read in the same scale, and I don't prevent them from playing with exagerating the values, nor make some of these values mutually exclusive.
Last, but absolutely not least, the Fcurves of opposing movments will have an harmonic flow that good animators with a solid visual perception can tweak for inbetweens without touching the blocking, pure curve work is a must-have for many technically savy (as in animation techniques) animators.

you have to think of combination sculpting as a sculpting system where people can individually tweak the contributions, and the hidden combinative balance will take care of blending these shapes and swap them around depedning on the channels' synergies.


But I do agree and feel the main reason is the animator would want complete total over what he has to work with. sometimes with joysticks you do play/fight alittle with joystick trying alittle too hard to get that inbetween bit " just over there but with alittle from back there sort of thing ":curious: in that case i jump in and edit a morphs or move a face bone But you can't do that at that level so Slider rule it.


as I said you aren't wrong, but consider that when working on feature films animators playing with a rig are usually tarred and feathered ;)

sticks have one nice thing to them, which is that their motion in their life panel (if your work is well laid out) will be representative of the arcs and movments in your rig, where a combinative system based on sliders will give you poses and precision you can't dream of with sticks, but the motion will need to be keyframed heavily or analyzed many times before you'll get the arcs and nuances of motion right, as what you are really keying are a set of very complex combinations, and not the motion of driving elements.

in feature films though you often go for this because, if a sliders heavy system can be made to move as well as a stick one with some overhead, you simply can't get the same poses and nuances with a stick based one.
you give up a bit of time-saving in animating motion, but you get a huge return in pure quality.

hope this helps shedding some light on the differences and choices between the two systems, I always find it funny that many people who do joystick setups cannot put the finger on the fact that the main argument in favour of sticks is this direct relation to the facial features motion :)

edit:
had forgotten the bit about the overhead of one translation layer in the first post.

T4D
07-05-2005, 03:51 AM
again Thanks :applause:

It sounds like Features film rigging is all about complete control over everything.
( not that I didn't think that already :D )

at that level character animation is at the level of keyframing Everthing at every key so sliders do make it the way to go.

tho not many have that amount of time to spend on CA
SO I still love my Joystick rigs they fun & fast :buttrock:


yes I would admit Joystick visual interface make it better for me, and most i do it for love'em too.

ThE_JacO
07-05-2005, 03:59 AM
It sounds like Features film rigging is all about complete control over everything.
( not that I didn't think that already :D )


sometimes it is, like on big characters, and that's where it's both fun and incredibly tedious at the same time, some other times it's about fudging things together in no time, some other times again a balance between the two :)


at that level character animation is at the level of keyframing Everthing at every key so sliders do make it the way to go.

tho not many have that amount of time to spend on CA
SO I still love my Joystick rigs they fun & fast :buttrock:


yes I would admit Joystick visual interface make it better for me, and most i do it for love'em too.

actually the two things are not mutually exclusive.
I believe part of the immediate future and present of rigging is about mixing the techniques.

things like mandibles and animal ears are still best off with bones and deformers and on-screen control, as the resulting changes in shape and silouette are massive and the feedback immediate.

some things, like subtle nuances, are best left to absolute control, but this doesn't mean the middle, and more motion dependant, elements can't be left to a good on-screen control system of some kind.

I believe the way to go is to use familiar interfaces and blending them together to optimize as much as possible screen estate and immediacy of controls.

an old tech soup I had always said: "try new things, so that you'll know what they're about when they'll be old and useful"

T4D
07-05-2005, 04:47 AM
your a:bowdown:,love to buy you a:beer: Yeah ! :buttrock:.

RenisanceX
07-06-2005, 04:01 AM
Wow thanx jaco i really

CGTalk Moderation
07-06-2005, 04:01 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.