VFX - Is it as Simple as it looks?
We're starting to get more and more visual effects work at Spin The Yarn lately, so we thought it would be a great opportunity to blog about it. I wanted to talk about the simplicity of it in particular, as the amount of work involved doesn't often equate to what you see in the final product. So, is VFX as simple as it looks? It's a bit of a loaded question, but I'll do my best to give my take on it. If you're not sure what VFX is exactly, I wrote about the differences between VFX, Motion Graphics and Animation here.
I'll start off by tackling the word "simple". When we hear simple, I think most of us generally associate it with quick and easy. In the world of motion graphics, when clients ask for "something quite simple", they're often after a clean minimalistic style. Not knowing the work involved, they could quite easily assume that it'll be quicker and therefore cheaper to produce, as there's less being shown on screen. This reminds me of a letter Winston Churchill once wrote, signing off with:
"If I Had More Time I Would Have Written A Shorter Letter"
It takes time to strip away what's not needed, leaving something that gets the job done while looking good. How does this relate to VFX? VFX is in essence re-creating reality. It doesn't necissarily have to be our reality. It could be a reality with magic, or giant robots, or giant magic robots. The VFX elements, no matter how fantastical, need to look real within the construct of that reality. In essence, the VFX artist is re-creating what the footage would look like if the visual effect was real and in the shot during filming. When making a giant magic robot look realistic, you can often hide behind some visual flair to sell the shot. This could be particle systems, lens flares, or maybe the robot is fighting another giant robot, giving us lots of lovely motion blur to hide any shortcomings. Creating a simple VFX shot is often quite difficult. You don't have anywhere to hide. The image above, at a glance, looks quite simple. It's a man wearing some very hipster looking glasses. As you may have already figured out, it's not quite that simple.
By looking under the hood, we unveil the level of complexity that's involved in creating something that on the surface looks quite simple. As this is footage, we need to make sure the glasses stick to the face. To do this, we track several points on the face. The tracking software can then calculate the position of each point in 3D space, allowing the software to create the combined movement of the real camera and the actors head. We want the glasses to cast a shadow on the face too (Look at the "Glasses Render" image above. Without the shadow, the glasses look very fake). To achieve this, we roughly model the shape of the face around the nose. It doesn't have to be perfect recreation of the face, as long as the shadow looks good, that's all that matters. This face model is called a shadow catcher, as it catches and only displays the shadows cast upon it. Both the Shadow Catcher and the CG Glasses stick to the face, as the CG camera is the only thing that's actually moving in the scene.
It's too obvious to say, but I'll say it anyway... Glasses are transparent. Because of this, we need to see the actors face through the lenses. We have the footage as a 2D plane moving along with the camera. This footage is emitting colour information through the glasses, adding an extra level of realism again. I modelled the glass to have thickness and a curved shape. The glass material has an index of refraction (IOR) that bends light. All of these work together in the final shot to distort the eyes as they would through a strong pair of glasses. Again, another little touch to push us towards realism.
All of the work up to this point has been done in Blender, an open source 3D application (it's our workhorse for most projects). Once we're happy with how everything is looking inside of Blender, we render everything out for some final tweaks. Each element is rendered out on a seperate pass, allowing us to tweak each individually. These passes are all brought together through compositing. This is where we will push the realism of the VFX a couple of notches futher again.
For this piece, we used Natron, an open source compositing application. Natron is node based, meaning that compositing is a case of stringing together boxes (nodes) to get your final result. Each node does something different and the image flows from top to bottom of the node tree. You can see near the top left of the tree we have the footage. To the right of this, we have what appears to be a black node. This node contains all of the CG elements rendered out of Blender in the form of an openEXR multilayer file. The Merge nodes merge two images togther. On the left of the node tree, we have two merge nodes below the footage node. These are merging the shadow and the ambient occlusion with the footage. We added a bit of a blur to the shadow node (the orange node near the top). Further down the node tree, we have another orange blur node, leading into a colour correction node, which then merges back into the main image. These blur and colour correct nodes are tweaking only the glasses. Compared to the footage, the render of the glasses out of Blender were way too sharp. Adding a slight blur helped it blend into the image. The colour correct node allowed me to darken the glasses and drop the saturation slightly, again, to help make them blend into the image. The yellow nodes at the bottom are writing out the final result, which in this case were ProRes Quicktime files. We have some more nodes dotted around that were used for saving different images and passes for this blog post. Long story short, the renders out of Blender were in the ballpark of what we were after. Compositing takes us to the pitchers mound.
After we finish compositing, we have the final shot. Here's a video showing all of the different elements in motion.
Hollywood Blockbuster VFX Porn
When most people hear that a film is VFX heavy, their thoughts most likely jump to the likes of the Transformers films. These films are littered with so much over the top VFX that they're often referred to as VFX porn. The effects are so obvious and so frequent, it's unavoidable to notice they are present. In 2014 the most VFX heavy film of the year was Birdman. This may sound surprising, as you can't really see the VFX most of the time. In my opinion, this is VFX at its peak. It allows the story to be told in the way the director intended without distracting the audience from the story. The film follows an actor during the leadup to his theatre production. There are no cuts, the camera flows around the environment, following the story as it unfolds. This is known as a "one shot" film, albeit a fabricated one in this case. It would have been impractical to film this as an authentic "one shot", so some VFX magic was used to stitch all of the shots together seamlessly. If you haven't seen Birdman, I very highly recommend it.
Simple is hard! Well, not too hard, but as far as VFX goes, there's a lot of work involved in making a simple VFX shot. We put the glasses sample together in about half a day, but if you want to see some of our more substantial VFX work, have a look here.