Monday, February 11, 2013
Trollhunter and Cloverfield VFX breakdowns
I found a few videos giving an insight into the development of the models and VFX techniques used in the film Trollhunter. The film is a "mockumentary", shot in a similar style to The Blair Witch Project; although it is arguably an overused cinematic technique, I thought it created an intriguing atmosphere in the film. Seeing these massive, grotesque creatures through the lens of a camera being held by a character makes them seem much more intimidating and scary.
The below video in particular gives details about the structure of the models and the different effects applied to the video. Muscle simulation was applied to the rig to create more organic movements of the mesh. Additional 3D details like growths and plants are then applied on top. It's clear to see that a great deal of work is put into post-production to ensure the monster fits the natural environment; the footage of the man on top of the jeep is in front of a green screen, the sky, troll and other environmental elements are dropped in, and many techniques and effects are added to emulate the "handheld footage" feel: matte separation, lens distortion, camera shake, grain match and colour correction.
The video also shows a few of the steps taken to create the scene of the monster being viewed through the jeep window; the troll's wiremesh is placed in the correct area, occlusion effects are applied to approximate realistic lighting behaviour and then VFX elements, like snow, are added last.
There are many different aspects to the rendering of the troll. As listed in the video, these include the beauty pass, vegetation pass, volume light pass, blast light pass, fur ambient pass, fur key light pass, ambient pass, soft reflection pass (and that's not all of them).
I don't know much at all about this side of visual effects and model rendering; it all seems way over my head at the minute. I did consider experimenting with muscle simulation in my model, but seeing as the focus of this module is on compositing and post-production work, I don't want to create much more work for myself than I already have.
I also found a breakdown of the effects used in Cloverfield, which similarly is through the point of view of a character and their camera. Here we can see techniques such as virtual set extension, (which I was surprised to discover a while ago is used widely in TV series so that they don't have to shoot on location). 3D elements include the head of the statue of liberty, bridges, aliens, etc. which are all composited with background and sometimes foreground plates to create the final clip. It seems the aliens are animated within a 3D structure reflecting the environment of the film to dictate where parts of it might be hidden from view, or where it interacts and collides with the surrounding buildings.
It's very clear to me that much of what makes a model convincing is how well it fits into it's environment, in terms of aspects like lighting and colour correction. The model I'm making for this module might not be anywhere near Hollywood standards, but I'm hoping if I spend a lot of time tweaking lighting and grading it won't look too separate from it's surroundings or from the actor.