Terminator Salvation :: Production Focus

[CryingHorn]

Awesome, thanks. I expected to know more in depth about cgi Arnold which in my opinion looked outstanding. After I saw t4, I came back home, found one t4 trailer which showed Arnold, and analyzed it for like 20 minutes with my jaw still dropped low. A huge respect to ILM.

[gjpetch]

I don't quite get sentences like:
"(the renderman shader) captured new subtleties like the bending of light around curves which could be used during animation."
Either that sentence doesn't make any sense, or I'm too tired.

[jojo1975]

Kudos as always to the masters of ILM. Actually I found it a very good movie and CG was very well integrated in the whole film.
Even if it is fish... it shouldn't smell like fish ;)

[richardmanhammer]

the article is decent, but there is maybe a sentence or two about Arnold.

[polygun]

the article is decent, but there is maybe a sentence or two about Arnold.


Yeah, barely anything about the 3d t800 endo either, im making one for fun and i was hoping for some insight

[vfx]

Thanks fot the article... but I have to say, it's becoming a bit fustrating these days reading half a dozen vfx articles that all cover the exact same topics. It would be so much better if the fx companies would give a little more insight into the work they did on specific shots rather than some broad descriptions here and there. I've read about 4 articles on Terminator and they all gloss over the same bloody things... shame cos I am willing to read more, for example how they pulled off that half terminator attack in the helicopter - that was so good and was really grounded in reality.

This same thing has started to plague Cinefex.

[Gpedder]

This part on page 3 at the top makes it sound as if ILM are the writers of PIXAR's Prman, I'd probably change it, and bending of light around curves sounds a bit silly too without any examples of why it was needed.

[Syndicate]

I was pretty let down by this movie. I wont go on about t2 being better or anything, however I think there were simply too many effects rather than just a few AMAZING effects.

By the time I saw CG Arnie, I was simply not that excited. I felt worn out as though I have just been watching a whole series of transformers.

Future battles in Terminator looks amazing in the other movies. What the heck is it with day time? and why why why did they lower the rating so much. Kids were in the cinema and constantly asking questions to their dads.

We need a director with a Midas touch, not the other way around.

I thought the trailer was kickass though :) the NIN version. Man that was more entertaining then the entire movie.

And yeah the article needs more beef. ;)

[kelgy]

After I saw t4, I came back home, found one t4 trailer which showed Arnold, and analyzed it for like 20 minutes with my jaw still dropped low. A huge respect to ILM.

**hey do you have a link to that particular trailer?

[patient09]

good read.

thanks.

[ThE_JacO]

This part on page 3 at the top makes it sound as if ILM are the writers of PIXAR's Prman
If you go far enough back in time... ;)

I'd probably change it, and bending of light around curves sounds a bit silly too without any examples of why it was needed.
That had me a bit puzzled too.

[MisterE]

**hey do you have a link to that particular trailer?

Here ya go:
http://www.youtube.com/watch?v=CQF3HEQwq2c

Don't think there's an HD version, though, sadly :(

[Ammarkk]

Very Good article...Thanks alot Paul.:love:

[Lman]

seen the film last night, really great work :bowdown:

A good blend with the live action and a good over all balance to the visuals.

[Gräck]

Terminator 4 is the most boring film I've ever seen. It was the first time I was close to leaving the cinema hall before the film ended. And it was also the first time I have not spent a thought about a film afterwards. I had no expectation before and though I was very disappointed. It was just stupid action, not really good action because there was no suspense. The flair of the previous terminator films was completely missing.

Sure, the visual quality was excellent.

[masterofacid]

Here is the complete Arnie clip!!!

http://www.youtube.com/watch?v=aY57vJOQIlE&feature=related

[Navstar]

Another summer movie filled with AAA VFX, a terrible script, and characters I couldn't care less about.

All those people doing all that great work on the CG... and the film gets a big 'meh'.

[phobos]

Thanks fot the article... but I have to say, it's becoming a bit fustrating these days reading half a dozen vfx articles that all cover the exact same topics. It would be so much better if the fx companies would give a little more insight into the work they did on specific shots rather than some broad descriptions here and there. I've read about 4 articles on Terminator and they all gloss over the same bloody things... shame cos I am willing to read more, for example how they pulled off that half terminator attack in the helicopter - that was so good and was really grounded in reality.

This same thing has started to plague Cinefex.

To feed your hunger you have to buy one magazine. Cinefex. It's a whole magazine dedicated to effects. Each issue comments on 3-4 movies which means each gets at least 20 pages of material.
It's not as good as it was 10 years ago but still miles ahead from the 5 page features of other magazines.


EDIT: I seem to have missed your last sentence where you clearly mention Cinefex!!
Yeah it's a shame the quality isn't as high as before.
If you can get your hands on it, you have to read the T2 issue of cinefex (no47).
It's like night and day!! Well written with nice trivia. In these days they seem to have a checklist. Talk about this cg scene, check, that cg scene, check etc.
Before it was all about the effects, not just cg

[SheepFactory]

Really dig those breakdown videos and hope to see that in future articles as well.

[eXcrem]

Awesome!! its very nice 3d mod terminator"

[gerardo]

I don't quite get sentences like:
"(the renderman shader) captured new subtleties like the bending of light around curves which could be used during animation."
Either that sentence doesn't make any sense, or I'm too tired.

This part on page 3 at the top makes it sound as if ILM are the writers of PIXAR's Prman, I'd probably change it, and bending of light around curves sounds a bit silly too without any examples of why it was needed.

As far as I understand from the article, they developed (for IronMan movie) a shader software for RenderMan that they has customized for T4 in order to match the real reflected light of detailed metallic surfaces in the animatronics. According to it, this software is able to capture/represent (it seems at shading level) new subtleties like the bending of light around curves. This 'bending of light' effect can be simply understood as the light reflected from a curved metallic surface, that is to say, caustics. This is an overlooked effect, mostly in light reflected from a metallic surface to the same metallic surface (as in the intricate and concave shapes of that kind of characters). Nor an easy effect (free-flickering) for animation.



Gerardo

[burnetrhoades]

After reading both this and other articles on the lighting technology used for the film I'm also left scratching my head why they seem to be making a big deal over techniques that aren't new and now almost standard procedure at shops all over the industry, in some form or fashion.

I enjoyed the film and liked the look of the robots a lot, they're my favorite aspect of the film, but image based lighting techniques have been used for nearly a decade in motion pictures. In my own work the use of specular() in any of my shaders has been verboten since the mid 1990s. The articles I've read seem to be implying that not only is an image based approximation of physically based shading something new but high dynamic range rendering as well, developed only as far back as ILM's work on Iron Man.

They may not be implying this at all but that's how I read it. Comments have already been made in this thread on the current state of technical journalism as it relates to CGI and effects. It's been bad for a long time. This confusion could be more to do with the understanding of the writer rather than what anyone at ILM is saying.

I had to applaud when someone referenced the Terminator 2 issue of Cinefex, the first issue I ever picked up and read, as being very different than what goes into contemporary articles. I couldn't agree more. I stopped relying on industry publications for actual insight into techniques and technology a long time ago. Word of mouth and intuition have both served me much better for a good while.

As far as I know the first feature film to make use of techniques based on the image based lighting research of Paul Debevec, confirmed by him at the time, was my own approach to the lighting of Senator Kelly for his death sequence in X-Men which was in theatres nine years ago. Both he and I presented on the application for the film at the following Siggraph in New Orleans.

Using geometry to represent a light source instead of relying on the hackery of specular/phong lighting tricks goes back much further than this. I believe I was first introduced to the technique by Jay Riddle in 1994 while he supervised my work on a Mazda commercial at Digital Domain.

[gerardo]

When a technique gets improved to a new level by an innovative approach, it's indeed a big deal, I think. I've had this experience some time ago with IBL techniques after designing and implementing a new CM workflow for my CG work (which also includes a DI proxy). This indeed changes the procedures and final results because we don't work with small color spaces. The size and shape of the color space is as relevant as its dynamic range, and this is a very important aspect very overlooked. Difference is that we replicate the real color flow for our CG elements and they integrates more accurately with real footage and we can even get more room for playing with color grading in the DI process.

I guess - just guessing :) - they are using a similar approach here (the same as, it seems, DD made with TCCOBB).

The luminous cards trick is indeed old too (it makes me remember in some way to the SIBL method) but what is interesting is the GL preview in the lighting setup from their LayerCake software. We may think this implies to translate the luminance and color data from a card into a lights rig with the proper intensity and color, over the animation.

The mix of diffuse/specular shading (from IBL rigs) and ambience/reflections is also very common, but this is commonly made by sight, and what it seems new for them is the energy conserving shader (which should be applying a balance between the combined usage of ambience/reflections and diffuse/specular automatically). This, mixed with the proper color flow for IBL indeed might put all these old techniques a bit further.



Gerardo

[playmesumch00ns]

Wow, really didn't learn anything from that article at all :\ The shot breakdowns are nice, but it's just showreel material.

And what the hell is 'bending light around curves' in the nonsensical blurb on 'renderman shader software customisation' supposed to mean?

Sorry guys, but VFXWorld's article on this was much better

[burnetrhoades]

Even the integration of CM tools to the lighting process has been done for some time, though it's still an evolving process. Sony Pictures Imageworks had a fairly advanced approach to this through the use of their own "socket renderer" module which replaced an application's built in Renderman driver, allowing the preview of any modeled colorspace right in the animation or lighting environment, pre-composite.

This goes back further than Spiderman 2 and Big Fish which was when I was introduced to it at SPI. Around the same time floating point linear light workflows were being integrated into the pipeline at Digital Domain and Rhythm and Hues. The GL preview aspect of ILM's technique is quite interesting and would represent a nice crossover of what might be considered advances in game shading techniques. There isn't enough effort being done these days to exploit the quantum leaps that have been made over the last decade in GPU performance and capability...in high end applications.

That's the rub and the irony. The drivers and GPU techniques for high end studios (due largely to the shift to the Linux platform, sadly) is quite limited compared to what (pro)/(con)sumers and gamers have been enjoying for some time. Besides IPR there is very little exploitation of realtime techniques thanks to floating point muscle of even disposable, commodity GPUs that dwarf the capabilities of our fastest multi-core, multi-processor workstation CPU configurations.

I was quite shocked and disappointed to learn that many of what seemed to be standard issue graphics for our latest and greatest and fastest "FX rig" or simulator's workstations housed pre-CUDA GPUs, precluding the integration of newly simplified accelerated physics solving into our workflow.

My hat's off to ILM for boldly going where few, curiously, are right now but, still, the rest of it has been around some time. I'm not wanting to take anything away from them because T4 was one of the first movies in a long time I've been truly excited to see, specifically because of the eye candy.

But I've observed a phenomenon at multiple facilities where the "lifers" can have a skewed perspective on what is happening in other parts of the world, that's all and what I'm guessing these recent technical articles may represent. ILM, and other studios, jockey for positions in various disciplines at the bleeding edge of what's being done all the time but aren't immune to being isolated, if you will, from techniques and technologies employed throughout the industry in other areas.

There was a time where they simultaneously employed an amazing optical-flow method for plate clean-up and artifact removal while an army of human match movers set up every single shot of every show because they had no 3D tracking software at all integrated into their workflow. But, I'll admit this could also be a case where they can't really talk about the truly neat stuff going on and their enthusiasm bubbles over into what they can talk about that's not really new, creating possible confusion in outside observers.

cheers

Sean

[soulburn3d]

Comments have already been made in this thread on the current state of technical journalism as it relates to CGI and effects. It's been bad for a long time.

Couldn't agree more. I still buy Cinefex, but it's now only for the pretty pictures. I got tired of reading articles where they say completely fabricated things. And certainly Cinefex isn't the only offender. It's such a shame. As someone working in the industry, I can poke a few friends and ask them for the real scoop (or at least the scoop from their perspective), but for someone in school or someone just getting interested in the field, it's sad that they are getting bogus information, and in all honesty it's harmful as well, since many of their decisions about what to study, etc are based on this info.

Makes me question as to whether magazine articles really were better in the old days, or if it only seems that way because I didn't know any better at the time. :)

- Neil

[playmesumch00ns]

It's not necessarily the journalists' fault though. I remember some ex-colleagues a few years ago joking about getting the stupidest name they could into cinefex for some imaginary new technique they developed. Unfortunately "hyperpolyvoxelnurbles" never made it in... :)

[playmesumch00ns]

This goes back further than Spiderman 2 and Big Fish which was when I was introduced to it at SPI. Around the same time floating point linear light workflows were being integrated into the pipeline at Digital Domain and Rhythm and Hues. The GL preview aspect of ILM's technique is quite interesting and would represent a nice crossover of what might be considered advances in game shading techniques. There isn't enough effort being done these days to exploit the quantum leaps that have been made over the last decade in GPU performance and capability...in high end applications.

That's the rub and the irony. The drivers and GPU techniques for high end studios (due largely to the shift to the Linux platform, sadly) is quite limited compared to what (pro)/(con)sumers and gamers have been enjoying for some time. Besides IPR there is very little exploitation of realtime techniques thanks to floating point muscle of even disposable, commodity GPUs that dwarf the capabilities of our fastest multi-core, multi-processor workstation CPU configurations.


Where was this GL preview thing mentioned? I didn't see anything about that in the article?

I'm not sure I follow why you think using a linux platform has anything to do with limited adoption of the GPU at high-end facilities? nVidia's linux drivers are pretty much on-par with their windows ones in terms of stability and features.

It's difficult to develop to take advantage of the GPU for several reasons - one is the constantly shifting goalposts of programming techniques required to actually leverage that power. Another is the exhorbitant cost of keeping up with the bleeding edge of that technology.

Ultimately you have to decide whether you want to spend your r&d budget on something that's only going to be used for previewing scenes or on software that's actually going to help generate the final frames. Since we get paid for final frames it's difficult to make a case for a massive spend on realtime technology.

[soulburn3d]

It's not necessarily the journalists' fault though. I remember some ex-colleagues a few years ago joking about getting the stupidest name they could into cinefex for some imaginary new technique they developed. Unfortunately "hyperpolyvoxelnurbles" never made it in... :)

True, their article is only as good as the info they're given. I know instances where a company has said they used proprietary software just so they sounded more hightech when in fact they used off the shelf software, and another company that said they used software X even though they didn't because they worked our a pr deal to get them a bunch of free licenses.

That said, I also know a lot of instances where a friend's words were totally rearranged and simplified to the point of being nonsensical by the journalist. So there's plenty of blame to spread around :)

- Neil

[TAVO]

When a technique gets improved to a new level by an innovative approach, it's indeed a big deal, I think. I've had this experience some time ago with IBL techniques after designing and implementing a new CM workflow for my CG work (which also includes a DI proxy). This indeed changes the procedures and final results because we don't work with small color spaces. The size and shape of the color space is as relevant as its dynamic range, and this is a very important aspect very overlooked. Difference is that we replicate the real color flow for our CG elements and they integrates more accurately with real footage and we can even get more room for playing with color grading in the DI process.

I guess - just guessing :) - they are using a similar approach here (the same as, it seems, DD made with TCCOBB).

The luminous cards trick is indeed old too (it makes me remember in some way to the SIBL method) but what is interesting is the GL preview in the lighting setup from their LayerCake software. We may think this implies to translate the luminance and color data from a card into a lights rig with the proper intensity and color, over the animation.

The mix of diffuse/specular shading (from IBL rigs) and ambience/reflections is also very common, but this is commonly made by sight, and what it seems new for them is the energy conserving shader (which should be applying a balance between the combined usage of ambience/reflections and diffuse/specular automatically). This, mixed with the proper color flow for IBL indeed might put all these old techniques a bit further.



Gerardo

Talking about nonsense......seriously, where do you learn all of this ?? Production ? College ?? the more i read about technical aspects of VFX productions the more i figure i do not know nothing about it.

I´m trying to make a move to VFX but all this information is way ahead my undertanding, do i really need all this education or it will come in time when need it ??. of course when working in production, if i ever do it.

[Navstar]

. Unfortunately "hyperpolyvoxelnurbles" never made it in... :)

That's pretty good! We made up a shader we called "Ambient Occuflection".

[burnetrhoades]

I'm not sure I follow why you think using a linux platform has anything to do with limited adoption of the GPU at high-end facilities? nVidia's linux drivers are pretty much on-par with their windows ones in terms of stability and features.
...

This is not true in my experience over the last nine years at three major houses (DD, SPI, R+H). Graphics drivers have not been, historically, on a par and as stable as their Windows counterparts. They have a better track record than ATI in this realm but it's still been a concern, big enough to sometimes influence which build of Linux gets used.

Don't think I'm advocating Windows. I'm just very, very disappointed after years of dealing with a dozen different flavors of pain-in-the-ass over the years.

[burnetrhoades]

It's not necessarily the journalists' fault though. I remember some ex-colleagues a few years ago joking about getting the stupidest name they could into cinefex for some imaginary new technique they developed. Unfortunately "hyperpolyvoxelnurbles" never made it in... :)


Hah-hah!

We used to do this in The Mansion for our procedural animation setups. Not to mess with journalists but in keeping with the tradition of rude software names at DD (nuke, hard, cloud bastard, voxel bitch, etc.)

Mathew Butler helped me write some auto-orient path routines to fly objects along motion paths that were actually drawn by feedback from keyframed channel motion with user scalable banking, drift, hover, etc. in Prisms. We called it Forward Unilateral Circular Kinematics. It actually entered the official team and dailies vernacular for a year or so later, applied to several shows. I know it made its way into at least one publication, maybe the DD coffee table book.

:)

[gerardo]

Talking about nonsense......seriously, where do you learn all of this ?? Production ? College ?? the more i read about technical aspects of VFX productions the more i figure i do not know nothing about it.

I´m trying to make a move to VFX but all this information is way ahead my undertanding, do i really need all this education or it will come in time when need it ??. of course when working in production, if i ever do it.
My friend, the learning is not a place, it's a volition of our mind :)

--- o ---

About the CM tools, yes, some few studios have been doing this for some time (I'd say more recently for some others), but the thing is that what they basically do is to work in smaller working color spaces (monitor) and later, when images are generated, they extrapolate these values to bigger cs. The problem with this approach is that the color flow is broken because they are beginning from their monitor cs, and the proper hues can't be recovered from a small cs. They can match the tones (because they are working in floating point), but they can't match the appropriate hues relationships, and the final result looks artificial in some way. It seems, by the results, they are solving these issues recently.

It's interesting to hear that their GL preview is more advanced, but not good to realize GPU techniques are still quite limited, and that's correct in general for all commercial 3D packages. It's impressive for example what an engine like CryEngine can do these days in regarding to lighting, shading, rendering and real physics. It seems Pixar is the studio that is taking more advantages of GPU techniques nowadays.



Gerardo

[burnetrhoades]

Um, no, the companies I mentioned are not working in a limited color space or generating original imagery in a limited color space and then extrapolating in a linear float working environment.

[gerardo]

Just in case - color space (cs) is commonly understood at gamma level only - I was referring at a gamut level. In the cases that I've seen, a studio may work in linear FP from the beginning, but the main limitation is the size (and shape) of their linear working color spaces for CG work. even when their CG pipeline is linear FP all the time, if the size of the cs is quite small compared with the film negative cs, it could bring some inconsistencies for integrations with real footage at the end (which they commonly try to solve in the DI process). The other issue is the setup (and tools) for proper color flows that should be hidden (as much as is possible) for CG artists. But it seems these things are evolving quickly. It's good to know that some few studios have these workflows already solved.



Gerardo

[ThE_JacO]

It's not necessarily the journalists' fault though. I remember some ex-colleagues a few years ago joking about getting the stupidest name they could into cinefex for some imaginary new technique they developed. Unfortunately "hyperpolyvoxelnurbles" never made it in... :)
I feel comforted by the fact you couldn't puzzle it out either :)
I'm not on the bleeding edge of rendering rnd, and am probably not going to go back there anytime soon, but to me the description has a strong marketing feeling with little consideration for that it's trying to describe.
It might be what Gerardo mentioned, but even then it sounds to me like it could be summed up into a more honest "implementing an evolved BRDF model suitable for metals", which would still sound suitably overcomplicated for an article :)

[playmesumch00ns]

This is not true in my experience over the last nine years at three major houses (DD, SPI, R+H). Graphics drivers have not been, historically, on a par and as stable as their Windows counterparts. They have a better track record than ATI in this realm but it's still been a concern, big enough to sometimes influence which build of Linux gets used.

Don't think I'm advocating Windows. I'm just very, very disappointed after years of dealing with a dozen different flavors of pain-in-the-ass over the years.

That's interesting. My experience has been different - I've never really had a problem with nVidia's linux drivers at all. The only niggle I can remember was having to wait something like a month for the latest beta drivers after the windows ones were released, and I only wanted them to get compiz running on my home ubuntu setup :)

[ianjarvis]

This thread is more interesting than the article.

I stopped getting cinefex as I actually began to find it a chore to read. Most of these sorts of articles have become rather generic, telling people the same thing with differently arranged words. Be nice for those of us that have a deeper interest than the average viewer to get a bit more depth or focus on the more unique aspects for different projects.

With regards to T4, I thought it was lacklustre. For me it is the final nail in the coffin for the Terminator series. It's just gone too far now from what one would hope a James Cameron directed followup might be like. The score was bloody awful too.

Back to the subject of vfx production, surely it is worth the effort to implement GPU workflows because of increase in productivity it might enable. Mudbox is a nice example of some software actually taking advantage of the GPU, and its nice that Adobe finally implemented gpu accelaration for PS CS4, although I've not been able to try it myself.

[Navstar]

This is a few days old, but another good T4 article from the Houdini guys...

They show the T-600 originally had a rubber face (as mentioned in T1)

http://www.sidefx.com/index.php?option=com_content&task=view&id=1539&Itemid=68

[Sandstorm]

I was really impressed by the atmosphere. This post-apocalyptic theme, the colors and the CG were great to watch. This was the first time I really saw no difference between the reality and CG. Funny, even the harvester had an exhaust system producing smoke. Just amazing. I hope to see more of this in the next upcoming Terminator movies!

Best regards,
Martin

[CGTalk Moderation]

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.