sebastian stan – Page 5 – knitting in the dark

xandraj5te:

The location for the shoot was in Switzerland, around Sion. The shoot was divided in 2 groups: the aerial Unit for the helicopter plates shot on Panavision Genesis and a ground unit with a tracking vehicle mounted with 2 Arri Alexa cameras on an Ultimate Arm. Dan also took thousands of tiled photographs of the environments and rock textures on canon 1Ds in order to build digital versions of the landscape.

We built the train model with an engine, 5 carriages and a caboose with gun turrets. This was then rigged with controls for the speed, the amount of movement between carriages and the banking. A separate rig allowed the artists to generate train tracks procedurally and constrain the train to the tracks. All this was animated in autodesk maya 2011. The textures of the train had to be seen fairly close. This meant having to use 3 high resolution textures of a resolution of 8000 pixels per carriage. In total, the train had 9 textures per channel (colour, specular, reflection, dirt, displacement, bump). Then each carriages had variations in the textures. This accounted for a total of over 80 textures. The texturing was done in Photoshop and Mari from the Foundry. The look development, lighting and rendering was done by David Mucci with Pixar renderman using HDRI image based lighting and raytracing. DNeg’s team also modelled, textured, and rigged the cable and zipline. The digi-doubles of Captain America, Bucky and Gabe were used in some shots at a maximum of a quarter of the screen height in pixels. They were constrained to the cable rig in a hanging position and were animated to the correct speed to land on the train.

To plan the whole sequence, production provided DNeg with a post-viz animation cut done by the Third Floor.We then had to model the whole environment based on the layout of the shots to ensure geographic continuity in the sequence. This meant having to go from large vistas to hugging a cliff side when on top of the train. This was a challenge in itself as the resolution of the rock face needed to holdup to full screen with a train going past at 90 mph, therefore covering lots of ground in a single shot.

To this end myself and our in-house surveyor, Craig Crane, took our lidar out to Cheddar Gorge, surveyed a number of locations to produce a vast high resolution mesh. This was then handed over to Rhys Salcombe to be cleaned in 3D coat and textured in Mari using a projection technique. The photography was sourced from the rock faces corresponding to the lidar scans. Rhys also modeled and textured a couple of viaducts that we see at the start and end of the sequence. The entire landscape was recreated in maya, with the addition of trees generated in houdini. Finally, the snow was added by a procedural shader in prman. For the distant plates, we used a mix of matte painting projection and the background plates shot in the alps. Various layers of effects and atmospherics were also added by Howard Margolius and his FX team: snow falling, snow being kicked by the train, mist and clouds and one hero explosion when a hole gets blasted in the side of the train.

from CAPTAIN AMERICA: Charlie Noble – VFX Supervisor – Double Negative October 4th 2011