you are here [x]: Scarlet Star Studios > the Scarlet Letters > lsgl: layers upon layers
<< before lsgl: the heat is on
after >> trixie's summer vacation

July 12, 2007

lsgl: layers upon layers

by sven at 11:59 pm

I thought that I'd completed the most complex shot in the film. ...I should know better than to say such things!

completed shot: explorers amid stampede
click on image to play clip (1.1 MB)

This shot has 20 Elder Things in it. The file was so huge, I had a 5 minute lag time just to select an item in the scene -- forget about modifying it! So, by necessity, I broke the shot up into parts.

I wound up having to do four separate renders out of LightWave. Let me walk you through 'em...

[Huh. Now that I think about it, I realize that it actually took me less time to render 20 elders using 4 passes than it did to render out 15 in 2 passes: 7 hrs 15 min vs. ~13 hrs. Interesting! It looks like there's a tipping-point at which managing too many calculations gums up the machine's ability to process them. ...Too bad that doing multiple renders requires babysitting the computer. I'll have to look into this further.]

backdrop - cave
click on image to play clip (48 KB)

This is what the cave looks like before I color-shift and hyper-saturate it in AfterEffects. Early on I tried to create a surface texture within LightWave that was this vivid -- but couldn't do it.

The shadows belong to the midground Elders. The models are actually in this scene -- I've just made them invisible.

I figured that the shadows of the background Elders would mostly be occluded, so there was no point in wasting render time on them. The shadows of the foreground Elders, on the other hand, would have fallen directly across the explorers -- so I decided to cheat the shot and have them not casting shadows either -- for the sake of a good visual composition.

Render time: ~165 min.

background - 9 Elder Things
click on image to play clip (32 KB)

I kind of like how the background Elder Things turned out... I'm wondering if I could recycle this animation during the "deep history" segment (Act II)?

Render time: ~117 min.

midground - 9 Elder Things
click on image to play clip (52 KB)

Once again, I did choreography using lo-poly stand-ins. Choreography took about 5 hours. Programming the choreography into the animated models took about another 2 hours. Doing the initial set-up -- matching the virtual camera angle to the real world camera angle -- that was ~2 hours, too.

In the midground, I put one Elder that's going faster than all the others to help create the feeling that these clones are actually unique Things. There's also an Elder that's about to pass really close to Andrew -- which sets up the next three shots, where Andrew's getting knocked down.

Render time: ~150 min.

focal point - explorers
click on image to play clip (81 KB)

The explorers are done entirely in AfterEffects -- which means essentially no render time to wait out.

foreground - 2 Elder Things
click on image to play clip (30 KB)

The front two Elders were the quickest to render...

Render time: ~30 min.

...After I had these five layers of animation, I composited them all together in AfterEffects. Color correction happens at this stage: I revved up the cave, and gave each layer a green tint. Actually, there are two shades of green going on -- "spring" and "seafoam." I've felt that simply tinting the room with one color gives too flat a result. (I'm lovin' what you can do with luma keying! Maybe I'll write about that later.)

Doing a depth-of-field effect in LightWave is extremely render-intensive. To simulate it "for cheap," in AfterEffects I applied a mild gaussian blur to the foreground layer.

Having a foreground layer hugely helps glue the explorers into their environment. Yet, when I watched a test clip, they still didn't feel as in-the-world as I wanted. So, I used a two duplicate layers of the cave backdrop that were brightened and masked off with feathered ovals to simulate gentle spotlights on Carl and Andrew. ...It's an effect that I've used "judiciously" rather than consistently: a suggestion that the "lavamen" actually emit light onto their environment.

Compositing time: ~1.5 hrs.

During the renders, I could usually go do other things. So... Amount of focussed attention on this shot: 13.5 hrs.

Amount of time the shot's on screen? Forty-eight frames -- about a second and a half.

But it's a really GREAT second-and-a-half, dammit! I'm proud of this shot. I think it's probably the best I've done.

posted by sven | July 12, 2007 11:59 PM | categories: let sleeping gods lie