Syriana, how did they do the explosion??

[XiaoSu Han]

There's this car explosion in Syriana,

(sorry for the tasteless music etc.)

 

my question is, how did they do it since you can see the people entering the car and it exploding without a cut?

 

did they just blow up some extras? :P

 

it also looks extremely real, so I don't know if it was CG?

 

Hope someone could enlighten me, regards, Xax

[Scott Fritzshall]

Two takes; one where the actor gets in, and one where the car explodes. Some tricky comping to get it to fit, since the camera is handheld and not motion control.

[Hristo Stoyanov]

I can't quite imagine it, can you correct me if i am wrong, please? You are saying that from the moment of closing the car doors - onwards: they reshot the scene, but this time without actors and with exploding car? The explosion shot how it was made - with a non-manned static camera, from a _best guess_ point of view based on the original camera path? And later how it was put together - They rotoed George to separate him apart from the BG explosion; masked everything from the first plate to reveal the explosion, then applied transformations to the explosion plate to match the shake/pans of the camera?

[Scott Fritzshall]

I couldn't tell you how they actually did it, but to me it looks like the easiest way would have been to shoot a take with just the actor going into the car first, with a relatively steady camera so as not to introduce additional motion blur, and then to shoot Clooney and the explosion after that. Then you've just got to roto out the other guy and the parts of the car that he interacts with, matchmove them to the plate with Clooney, and roto around whatever passes in front of them. There are fewer things to match that way, and less camera movement on the things that do. But again, I have no idea what they actually did.

[Steve McBride]

Go get the DVD for "Cloverfield" make sure it has the BTS and watch all of them. There is so much cool stuff that they show you in the vids it's amazing. That movie is a better example of CG in a handheld environment seeing as the whole movie was handheld and was basically all CG, I mean you can't really have a ten story baby monster going on a rampage through New York... or can you?

[Hal Smith]

Motion control is used quite a bit in the FX/CG world. It's quite possible they programmed a moco rig to look like handheld work so they could exactly repeat the moves.

[XiaoSu Han]

http://www.google.com/url?sa=t&source=...MQVRGBGBx7DZCyA

 

to keep it short: they did three passes with the dolly (moco I guess?) and then composited it together. nice.

[Chris Millar]

take note of the person walking in front of both clooney and the car door just as its closing ...

 

And yes motion control can do the movements you are seeing here

[Karel Bata]

Xiao, your pdf only shows up as a square here.

 

I would think: 1) conventional track and pan with Clooney shooting open gate, then lock camera dolly off 2) explosion 3) marry both shots together 4) add 'camera shake' later.

 

Looks simple enough... :lol:

 

What does your pdf say?

[Chris Millar]

I dont see what is so hard about this shot - sure it involves some tight timing and experienced operators (especially the pyro/explosives) but conceptually it aint that hard to come up with one or two ways to achieve it depending on if you want post or onsite intensive production. Always hard to be conclusive but I reckon the key aspect is the guy in the white shirt wiping across the shot from about 3.5 secs to 5 or so...

 

At a guess the 'explosion' is much more bark than bite, especially once sound designers get going and doing what they are paid for - the car rocking as it did was probably done with actuators...

 

Note not one flinch from Mr Clooney - indeterminate as a clue, but interesting ;)

 

The camera shake has parallax... cant do that in post (within a reasonable budget I should say)

[Karel Bata]

• 
  

This is from an entirely different angle. Rather odd if he wasn't there when they blew the car...

[XiaoSu Han]

Karel you might have to get a newer PDF reader, it works perfectly for me here...

 

it says that they used motion control and filmed 3 passes, one with Clooney, one with the explosion, one with extras... and then matting it together

[Karel Bata]

I can read it now. But it doesn't say an awful lot.

 

Do you think that still was composited later? There's some orange reflected on his left arm (well technically his right arm) that would be tricky to fake.

[Chris Millar]

• 
  

This is from an entirely different angle. Rather odd if he wasn't there when they blew the car...

 

Why is it odd if he wasn't there ? Notice that none of the other extras are there in the still ? Isn't that also odd ?

 

There are no rules to this except maybe sticking to a reasonable budget (which means time also) - other than that whatever works, go for it ...

 

Orange glow - howabout synchronizing some form of orange light pointing at him ?! Or yeh, he could be there - or, again - whatever works ;)

 

without seeing some sort of behind the scenes we will never really know - but again, I'll say there are a multitude of standard tricks and iterations of these to achieve this. Depending on available skill and location issues some will become more obvious choices than others, this is stuff that we cannot see in the frames presented to us.

[Scott Fritzshall]

Do you think that still was composited later?

Yup.

[Phil Rhodes]

Pretty clear jump cut behind the guy who crosses frame.

 

I'm going for moco, or moco with a bit of extra post wobble, with a cut hidden by the extra.

 

P

[Karel Bata]

Hm...

 

re the still: It's so obvious once you know, and makes me feel foolish for not seeing it at first. It just never occurred to me that a stills photographer might anticipate doing the same compositing trick and set up accordingly. With that compression of perspective he (or she! slapped wrist here) must have been way back with a long lens.

 

Chris: "The camera shake has parallax... cant do that in post ", then how was it done? I don't think Syriana had a huge budget. Or is even doing that cheap these days...? I clearly can't keep up! :unsure:

:D

 

Sorry Phil, what's moco?

Edited January 27, 2009 by Karel Bata

[Chris Millar]

moco - motion control

 

The parallax was done by wiggling the camera.

 

Phils suggestion of both moco and a post approach seems like a wise consideration to keep the budget down. Would be interesting to make an application for AE/Flame whatever that was smart and could recognize camera shake direction over time and augment it in an apparently intelligent fashion i.e. always working with it as opposed to the simple variations of random settings that most seem to have.

 

I still maintain however that todays automation technology could cope with such a motion path (if for whatever reason it were the only way to go about it)

 

If I were designing a system like that to avoid the power requirements and bulk of dealing with the momentum I'd have the standard moco rig do the main move and a smaller specialized system synced and attached to that like a head to do the repeat camera shake moves...

[Karel Bata]

Would be interesting to make an application for AE/Flame whatever that was smart and could recognize camera shake direction over time and augment it in an apparently intelligent fashion i.e. always working with it as opposed to the simple variations of random settings that most seem to have.

That's easy. I've done it several times. No application is required just a knowledge of coding. In AE you would create a composition and motion stabilize it, create another composition using the same footage, and write a function that took the data from the first, multiplied it, then added it to the second. Voila! You can tweak the degree of 'hand-held' as you go along.

 

To be more technical -you create 2 Expression sliders to hold the multiplier values, which would allow you to keyframe the percent of stabilization (on both axis). Basically, you need to calculate the difference between the stablilized value and the original value, multiply that by the sliders, and add that back to the original values.

 

And you certainly don't need to rely on any random settings. You can create a null object that tracks the movements of your mouse, then (while watching the footage) move that in a 'hand held' sort of way, and then apply it. You'd then tweak the key frames. Adjusting the 'skew' would be done separately, but it's a real pig because a small tilt to one side can cut off a lot of image area which then needs fixing.

 

Reminds me of how horrified I was when I first heard two 'creatives' on an ad discussing whether to make a shot hand held - they concluded they'd do it in post. Sacrilege! :o

[Chris Millar]

That's easy. I've done it several times. No application is required just a knowledge of coding. In AE you would create a composition and motion stabilize it, create another composition using the same footage, and write a function that took the data from the first, multiplied it, then added it to the second. Voila! You can tweak the degree of 'hand-held' as you go along.

 

Nice !

 

hmmm, with a bit of math/manipulation you could also take the moco data from the previz application you programmed the moves in (3DSmax, et al) and use that.

 

You could then program in the post wobble into the 3D previz/moco programming package that and it would tell you how much wider the shot needs to be to account for the frame edge falling into view ... (previz-ing post effects :rolleyes:)

 

Would be interesting to compare the output and time elements in workflow of a proactive vs. reactive approach.

[Karel Bata]

I'm still not convinced it's moco (hey - I've used a new word today!). Didn't they shoot that in Morocco. Getting a moco rig out there would be a hassle.

 

I'll be back...

[Scott Fritzshall]

I'm not convinced either; that PDF just says they did multiple passes on the dolly. That's why the compositing is a bit tricky; you would have to put together several passes that aren't identical. Most likely you would do a 3d camera track of all passes, recreate basic scene geometry, project the second and third plates onto that geometry, and rephotograph it through the primary camera (probably the pass with Clooney). Then a lot of roto to put the pieces together properly.

[Karel Bata]

Whoah! I really think it's a lot simpler than that - which is why I find it interesting. I'll be looking at the film over the weekend.

[Scott Fritzshall]

That would be the general process for syncing up elements from shots with similar but not identical 3d camera movement. It might have been simpler for this shot though. Still lots of roto for the people, and definitely some sort of matchmoving to get them to sit in the plate.

[Karel Bata]

What's interesting about analysing footage like this is that different people see different things. You saw the significance of the man crossing frame. I missed that. You saw it as hand held, but I see it as a track with wobble added - maybe because I've done that myself. What I felt certain I also saw was that the camera was entirely still during the explosion. So a closer look:

 

The original footage at a better quality:

 

And partially stablised:

I didn't spend a lot of time on it so a few bumps remain but it's enough. (BTW does this YT Video play ok for you? I've added a huge amount of black at the end to test something.)

 

It is indeed a track. Importantly, the camera is stationary at the crucial point. After the explosion, when the film's image pans off to the right of the YT window, you can briefly see that the explosion was a static shot and moved left off-screen in post to match the foreground pan. And (or is it my imagination?) the background zooms in slightly as the explosion happens.

 

I thought I saw some tilt, but stabilising in the x and y axes reveals there's none at all. Interesting illusion that.

 

No moco needed for this. I think it's very clever and a cheap way to do it. <_<

 

(The eagle-eyed might notice that all number plates are blurred out. :huh: )