Weta Digital's Making of – The Hobbit An Unexpected Journey - 3 Videos

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

THREAD CLOSED
 
Thread Tools Search this Thread Display Modes
Old 01 January 2013   #1
Weta Digital's Making of – The Hobbit An Unexpected Journey - 3 Videos

Weta Digital's Making of – The Hobbit An Unexpected Journey - 3 Videos
http://bit.ly/The_Hobbit_An_Unexpec...by_Weta_Digital


Making of – Gollum, Goblin King & Azog Videos

Making of – Gollum - http://bit.ly/Gollum_MakingOf
Making of – Azog - http://bit.ly/Azog_MakingOf
Making of – Goblin King - http://bit.ly/GoblinKing_MakingOf
 
Old 01 January 2013   #2
Comparing the old gollum with the new one is awful and interesting.
Many things have improved, not exactly technology, but approach..
Nice to see good use of saccades and blendshapes.

I wanted to know if they used a physically–based simulation for the faces.. or at least some, but the videos and articles, show only, that it was used mainly for the body.
Reliable facial animation, using physically–based methods, is still very far.
But the muscle rig + skin simulation, really helps and adds extra realism.
Though I sill see the bodies rigid in many parts, and the skin varies in realism depending on the amount and kind of motion. I don't know if it's mainly because of how it's rigged or limitations of the skin simulation.

'Tissue'. No technical paper, no more information about it, pros or limitations.
That's the bad thing, not much information.

The main secret for good results, I see, still doesn't depend, not on novel technology, but the detailed final work of all the people involved.

.

That site has edited the videos to add watermarks(bad),
I extracted ones with more quality and probably the originals from:
http://herocomplex.latimes.com/2013...ng-and-azog/#/0
__________________
may not be following this thread.
.
 
Old 01 January 2013   #3
Originally Posted by Ruramuq: Comparing the old gollum with the new one is awful and interesting.
Many things have improved, not exactly technology, but approach..
Nice to see good use of saccades and blendshapes.

I wanted to know if they used a physically–based simulation for the faces.. or at least some, but the videos and articles, show only, that it was used mainly for the body.
Reliable facial animation, using physically–based methods, is still very far.
But the muscle rig + skin simulation, really helps and adds extra realism.
Though I sill see the bodies rigid in many parts, and the skin varies in realism depending on the amount and kind of motion. I don't know if it's mainly because of how it's rigged or limitations of the skin simulation.

'Tissue'. No technical paper, no more information about it, pros or limitations.
That's the bad thing, not much information.

The main secret for good results, I see, still doesn't depend, not on novel technology, but the detailed final work of all the people involved.

.



That site has edited the videos to add watermarks(bad),
I extracted ones with more quality and probably the originals from:
http://herocomplex.latimes.com/2013...ng-and-azog/#/0



Totally awesome!,

From watching the video, they might be using PSD for there muscles - not because im psychic but rather thats what they named the file! (image attached)

And also because this file is a rom, one can speculate that it's a statistical model, for the final motion. I.e the final motion does a look up into the rom, and drives the PSD's accordingly.
Attached Images
File Type: png Picture 3.png (41.6 KB, 175 views)
__________________
Disclaimer: My opinions are not those of my employer.


 
Old 01 January 2013   #4
They said, on an interview, that the muscles were able to collide each other since avatar.
The video does not show that.. It has some problems too.
I'ts hard to say, how it really works. Or how accurate it is. We can only guess..
__________________
may not be following this thread.
.
 
Old 01 January 2013   #5
Originally Posted by Ruramuq: Many things have improved, not exactly technology, but approach..


The biggest differentiator is computer hardware. In 2002 it would not have been possible to work with a head that has 25-30 thousand faces, and it would not have been possible to load a facial rig with thousands of shapes into the memory of a 32-bit system; and the network components and harddisk capacity would not have been able to keep pace either.


Quote: I wanted to know if they used a physically–based simulation for the faces..


The problem with that is lack of interactive speeds.

It's not absolutely necessary to see the final skin when they're animating the body, but it is when working on the face. A muscle based system could probably not run even the old 2-3000 polygon Gollum head at acceptable speeds, and if it's not there the animator has no idea how the face looks like.

I would not expect to see a full muscle and soft tissue simulated face in at least another 5 years, maybe even more. It is not practical with current hardware speeds.

And even if the speed will be there... the current artistic approach seems to work rather fine already. The level of control that a combination blendshape system provides is very hard to do with sims, and the director won't always want a physically correct result anyway.
__________________
Tamas Varga
 
Old 01 January 2013   #6
Originally Posted by Laa-Yosh: The biggest differentiator is computer hardware. In 2002 it would not have been possible to work with a head that has 25-30 thousand faces, and it would not have been possible to load a facial rig with thousands of shapes into the memory of a 32-bit system; and the network components and harddisk capacity would not have been able to keep pace either.


Your saying that hardware and not people is what made the difference.
Somehow that underestimates the new work of the people involved this time. It's like saying that they could have done something similar, but they did not had the hardware.Absolutely No. that's ilogical.

We re not talking about massive amounts of information, etc. because Weta would have taken the best decisions depending on the situation, making sacrifices were necessary.
Besides the first gollum used blendshapes too. But were not based on muscles/anatomy like it is now(FACS).

There is a really big benefit in hardware. But it cannot be used to generalize the quality of the results.
Saying that it makes the biggest difference is very absurd.

There is new knowledge and experience they've got with anatomy/physiology of a character(and not just characters). Something they didn't really know before, as they said more than once in interviews. Also very noticeable in the Gollum of previous movies. That any experienced animator can notice.

They were exploring and testing without really knowing how muscles, skin really works. And that's something they have improved, probably since the day they decided to be more scientific.
They actually said something like that, that they want to do things based more on science.

Quote: The problem with that is lack of interactive speeds.

It's not absolutely necessary to see the final skin when they're animating the body, but it is when working on the face. A muscle based system could probably not run even the old 2-3000 polygon Gollum head at acceptable speeds, and if it's not there the animator has no idea how the face looks like.


'Interactivity' is ideal, but it's not the main problem. They use 'Finite Element' simulations, this means, they don't get interactive results anyways, be it face or body. It requires, who knows how many minutes per frame.

Weta, had/has people, with knowledge in bio-mechanics. And I'm sure they've tried, facial animation system based on that, and found that the results were not good, not because of the interactivity, but because of the complexity, that nobody is able to control yet. That's the real reason why they relied on their FACS system, because in the end it's much better, more detailed than simulations, though it depends on artists. And artists are not scientists. That's why the results may not be always correct. The previous gollum compared to the new one is a clear example of that.

I'm sure Weta is constantly trying, but it is not working and FACS has limitations , but works.
That's why I give a lot of credit, not to hardware, but the work of their sculptors/artists. Who have been very careful about several details of the face that was missing in the previous gollum. With or without new hardware.

A completely different approach since the previous LOTR movies.

Quote: I would not expect to see a full muscle and soft tissue simulated face in at least another 5 years, maybe even more. It is not practical with current hardware speeds.


It's not a popular topic. Most people working on it, are not doing animation.
The idea of using FACS for animation, is very old, 1981 or before, nonetheless it was mostly unknown in movies. Until recently, thanks to Avatar.

Despite of hardware, the math is simply not solved. For example Fascia is fascinating, yet not enough investigation. micro-wrinkles algorithms, not yet.

A 'complete model' for tissue simulation does not exist, but even if we use a 'basic simulation model', we just want it to be correct. We are not exactly there, but they are getting good results anyway.
There is also the case of different implementations, like face-robot tissue.. that does not require real-time simulation.

.

(It would be nice, more technical information about 'Tissue', to compare it with other researches.)

-

And I almost forgot, they changed the topology, and now use a much cleaner topology, which was unquestionably needed for more realistic animation:
__________________
may not be following this thread.
.
 
Old 01 January 2013   #7
Somehow I expected you to try to prove me wrong... I don't understand why, though, but let's go through it one by one.

Originally Posted by Ruramuq: Your saying that hardware and not people is what made the difference.
Somehow that underestimates the new work of the people involved this time.


Don't put words in my mouth please, what I said was that hardware development has enabled people to do things differently, and ultimately better.

Quote: It's like saying that they could have done something similar, but they did not had the hardware.Absolutely No. that's ilogical.


That is logical - Bay Raitt in 2002 did not have the computers to work with 30k polygon head models. He also did not have brush based sculpting tools as advanced as we have access to today that allow much faster production time.
Try to put together a Pentium 4 with a 32 bit Maya 4 or so and load up a 30000 polygon head model if you can, you'd be surprised how slow it would be. Try to UV map it without today's unwrapping tools, or work on it without Zbrush and soft selections. That's what they had back then, oh and Mirai too - the mother of all poly modeling paradigms and tools of today.
Even the Weta guys themselves have said that they just did not have the tools to do Avatar levels of stuff at that time.

So how could this reasoning be illogical?

Quote: Besides the first gollum used blendshapes too. But were not based on muscles/anatomy like it is now(FACS).


I'm sorry but you're as wrong as it can be. It was Bay Raitt's idea to build his blendshapes on FACS. It was also his idea to use corrective shapes for practically all the combinations of the individual AUs to fix the deformations. He came up with this whole thing on his own.
Have you ever read his posts on the Spiraloid forum? He openly told us all about his concepts and gave plenty of links and advice about it.

Quote: There is new knowledge and experience they've got with anatomy/physiology of a character(and not just characters). Something they didn't really know before, as they said more than once in interviews. Also very noticeable in the Gollum of previous movies. That any experienced animator can notice.


The first Gollum was breaking new ground and they were the first to use this system without any experience at all. The differences in how the animators were able to utilize the system between TTT and ROTK are just as noticeable. They have learned that much about the concept during the production of those movies.

And of course there's new knowledge, they did King Kong after LOTR and explored the concept even further. Then they did Avatar and learned even more. It's the nature of things, I've never argued with that.

What I didn't agree with is that there's some radically new approach. There isn't that much of a difference as it's still FACS based with corrective shapes - but now they have more than the original FACS set of Action Units, and they have more geometry to create the shapes, and lots of more advanced tools to manipulate the model more quickly. So the biggest difference between what it was possible to do in 2002 and in 2012 is the hardware (and somewhat the software). Then there's the experience gained from Gollum, Kong, Jake Sully and the rest.

Hell, if you watched the video on Gollum, how could you miss the original blendshapes displayed on the top left side when they go through the various expressions??

And this does not belittle the efforts on Hobbit at all. They've done some truly amazing work, they took a fantastic character and made it even better, and they've made a legion of new, equally fantastic characters too.
By the way, to me it seems more like you're the one with an agenda here, trying to make the first Gollum seem less like it was.

Quote: They were exploring and testing without really knowing how muscles, skin really works.


Again, Bay spent years researching this stuff for Gollum... Things like watching movies from the first row to focus on how the faces move, or stopping people for small talk just to see their facial expressions.
Then he tested different rigging approaches until he came up with this one because he thought it's the most solid.

Quote: They actually said something like that, that they want to do things based more on science.


That was probably about the muscle system.

Quote: Weta, had/has people, with knowledge in bio-mechanics. And I'm sure they've tried, facial animation system based on that, and found that the results were not good, not because of the interactivity, but because of the complexity, that nobody is able to control yet.


Interactivity is still necessary for animation and if they can't make a muscle sim based rig work fast enough then it doesn't matter how good the results are. The animators would just not be able to produce enough animation.

Quote: A completely different approach since the previous LOTR movies.


False again. Please do your reading if you don't believe me, there are some writeups about Bay's talks and such. Also, I'll try to ask him to see if he can put the forum back online, his site is working so he can probably do it if he has the files.

Quote: And I almost forgot, they changed the topology, and now use a much cleaner topology, which was unquestionably needed for more realistic animation:


Are you really serious about this? There are loops running in the exact same direction all over that head! Can't you see?



Here, red loops are the same, most likely the result of subdividing the original basemesh.

Green loops are a new development, Jeff Unay's contribution: when your mesh is dense enough, you don't have to follow surface details, you can just imply the shapes because you have enough vertices to do it.
Bay haven't had enough resolution so he had to follow the lines of the neck and the cheek folds with his loops. Once again the case of more powerful hardware enabling a new, more refined approach.

You can also see the FACS naming conventions on the left, although the text is very hard to read because it was captured from the LOTR DVD feature. I can find you other sources if you still don't believe me...
__________________
Tamas Varga

Last edited by Laa-Yosh : 01 January 2013 at 11:30 PM.
 
Old 01 January 2013   #8
nice one, Tamas
 
Old 01 January 2013   #9
Oh and another thing to add. All of the above should also make it clear just how much I - and probably a lot of other guys - have learned from the people at Weta, and how grateful we are that they were, and are, willing to share their knowledge and experience with us.
__________________
Tamas Varga
 
Old 01 January 2013   #10
I don't try, In fact I don't need to write much more.. because my argument has been always the same.
But you are going to insist, you want to be right at any cost.

Either displacement maps, (your previous argument)
Either it's hardware, (anything but not the old gollum topology)
Either, it's density, (look at the old gollum's density, )
Either, it's really the same topology, look, it's hidden, etc



--

Displacement :
Originally Posted by an interview: "With the wrinkles in the model, he could control the motion of them so that the skin actually squashes together and then forms the wrinkle, instead of it just dissolving on and off like a displacement."

They need better topology than before to support wrinkles properly. There is no hardware, displacement or any other justification. Good Topology is benefical, now or ten years ago!.

Old patterns found ?
Yes, what Laa-yoosh found, but forcibly, but omits the most important parts and the fact that some patters are considered generic.
What could possibly matter those patterns in the ears.
It's the same topology, can't you see(somebody said), but let's see:


-

Differences in topology:



And it's not even the topology itlsef, it's also the even distribution of poligons what matters, it's more anatomically correct.
Another very important fact, is that motion happens in a relative context, there is a continuity, that messy patterns can't possibly get. Not the most fancy/forced explanation can go against this, and it's importance in skin motion.

Something very important, is that the pattern in the cheek, is something that has been defended insistently by Laa-yosh , and that was the basis of long, and what I've always consider very misleading information about topology. But the truth is that now it's gone. they removed it.
You have to be very stubborn to go againt facts and it's not the first time.

density
Quote: when your mesh is dense enough, you don't have to follow surface details,

Something important about density, is that it's not something to brute force. no.
It can be calculated with math. This is done depending on motion and curve.
It is not an argument to support the old gollum model, which by the way has a very ba d density and distribution.
It's also not possible to use it as an argument because to support different characteristics of a faces, it would require a ridiculous density.


-- now some clarifications --

Quote: Don't put words in my mouth please, what I said was that hardware development has enabled people to do things differently, and ultimately better.


I don't think I did, i.e. here are your words :
Quote:
Quote: Originally Posted by Laa-Yosh
The biggest differentiator is computer hardware. In 2002 it would not have been possible to work with a head that has 25-30 thousand faces, and it would not have been possible to load a facial rig with thousands of shapes into the memory of a 32-bit system; and the network components and harddisk capacity would not have been able to keep pace either.


The biggest difference is always in the people. People makes tools, people learns, and tests new solutions.
They change, they pursue better quality, and for that, they learn, and if there is something wrong, they correct it.

Quote: That is logical - Bay Raitt in 2002 did not have the computers to work with 30k polygon head models. He also did not have brush based sculpting tools as advanced as we have access to today that allow much faster production time.
Try to put together a Pentium 4 with a 32 bit Maya 4 or so and load up a 30000 polygon head model if you can, you'd be surprised how slow it would be. Try to UV map it without today's unwrapping tools, or work on it without Zbrush and soft selections. That's what they had back then, oh and Mirai too - the mother of all poly modeling paradigms and tools of today.
Even the Weta guys themselves have said that they just did not have the tools to do Avatar levels of stuff at that time.

So how could this reasoning be illogical?

Again, It's ilogical to compare hardware in such literal sense in your post. Decisions are taken depending on the case.
It's also absurd to compare FACS too, knowing they are different systems.
Weta said many things in relation to hardware, and actors, and mocap system.
You are trying to be smart by mixing everything but we both know they refer to the whole complexity of Avatar. e.g. The performance capture of today allows things that were not possible before...etc.

Even if we ignore hardware. your argument does not stand correct, because the same methods of before would lead to very unsatisfactory results today. and viceversa.
(I better write the viceversa with words.. otherwise..)
The same methods of today applied to the tech of before would lead to really great results. with many limitations(hardware) but with a considerable benefit.

Quote: I'm sorry but you're as wrong as it can be. It was Bay Raitt's idea to build his blendshapes on FACS. It was also his idea to use corrective shapes for practically all the combinations of the individual AUs to fix the deformations. He came up with this whole thing on his own.


I already talk about this, this ideas of FACS are old, very old. I wrote about this in the previous post. And if you or anynbody things about FACS, then the relation with animation is not a mistery. It's not a discovery, or something novel, like you try to say.
FACS is something that has been in the head of many people. You like it or not.

"But I understand, that your admire, gollum modeling, and the person who did it."
Sadly I never heard of him before. And I don't use names or movies, as the argument. only what I read that i can prove with my own experience, not something I hear in a tutorial, video, talk, interview, rigging/animation theory, etc.

I can tell, with confidence, that people should not use previous gollum topology, Not. unless they are more like artists, with lack of experience in animation or technical issues.

Quote: There isn't that much of a difference as it's still FACS based with corrective shapes - but now they have more than the original FACS set of Action Units

No, that can't be true, are you speculating, do you have proves ?
The new system has to be different, they have to calibrate the system for mocap. I think there is a guy who won an award from the academy for the system of Avater.. Mark Sagar. if i'm correct.

Weta themselves have said many times how different things are than before and you say the biggest difference is hardware.
There are always similarities, always, but that's not the point.

--

I'm gonna say something very clear:
I don't have the most minimum interest to diminish the work done in the first lOTR, but I'm not going to act like a fanboy, who defends it's favorite movie. And does it with a new argument each time. With new explanations. despite the proves. Contradictory as I proved before.

I'm consistent, people can check everything I've said in the past.
It's simply not possible to explain the patterns of gollum, in a technical way, you can't write an algorithm with it. It's too arbitrary.

--
Now I see no reason to keep writing, because I mostly disagree with Laa-Yosh.
And I believe you have to prove yourself. like I did, really many years ago. And keep doing it.
(Even if you have to learn rigging or coding.. whatever..)
If you had a valid arguments. I'm sure I would follow your ideas to learn something truy new.

I see that you talk about things 'you've heard or read, or watched', that people says, and they are right,they, because I've read them too. But then you interpret them with your own criteria, despite your not a rigger, animator, coder. etc..

And I know you are going to continue until the end of the world. Giving arguments, and 10 years from know, you will find endless arguments, whatever the topic is. I now also remember very well, you getting very immadure in the topology research, and acting like a child.

And if you find that this post is a bit hard, I'm sorry because I don't really want to, but it's enough.
You don't even have the courage to admit your errors. Despite the proves.


--

ok, little by little I finished, My opinion is in there.
I let the last word to those who are always right no matter what. Or those who want to participate. I'm tired of this.


and metamesh, guess what, :] your participation in this topic, is the exact definition of a troll.
http://en.wikipedia.org/wiki/Troll_%28Internet%29
Instead of participating, like a normal user.. but maybe it's because your not a tech guy, I remember..http://forums.cgsociety.org/showthr...?f=98&t=1038637
__________________
may not be following this thread.
.
 
Old 01 January 2013   #11
**stares smiling at the thread while eating pop corn**
**crunch....crunch...crunch**

 
Old 01 January 2013   #12
alessandro please tease us with some insides...
__________________
...
 
Old 01 January 2013   #13
There has to be a more brief way to make points in these threads.
 
Old 01 January 2013   #14
I think I've managed to uncover a devilishly clever plan by a certain group of individuals...

Step 1
Make some awesome creatures for an awesome movie

Step 2
Release video with some - but not all! - details of how to do such creatures

Step 3
Make popcorn, lean back, and then watch fellow CG people stay up late thinking and arguing about the details that were intentionally left foggy

Veeeery clever indeed!
__________________
Tamas Varga
 
Old 01 January 2013   #15
Ruramuq

I've been thinking about it and realized two things

- you're running in circles, repeating the same arguments again and again, even though a lot of them are completely wrong, and a lot of them I've already answered

- continuing this pointless argument about your theories and ideas won't get me anywhere; but working more on our facerigs and thinking about what I can use from these videos to make them better will

So I've come to the conclusion that I just don't care about it anymore, I'm out and I leave you to believe whatever you want.
__________________
Tamas Varga
 
Thread Closed share thread



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
CGSociety
Society of Digital Artists
www.cgsociety.org

Powered by vBulletin
Copyright ©2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump
Miscellaneous

All times are GMT. The time now is 11:13 AM.


Powered by vBulletin
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.