Solar System & Breakdown from Liam Major on Vimeo
While filming project Small Boy, I recorded some hand held footage to experiment match moving with. The footage was quite grainy and had low levels of light and contrast, so I wanted to try colour correcting and blurring a pixel or two to see if it would improve the track, and it did. With this, I decided to turn it into a short project with a couple of weeks spent on it.
Figure I & II: Raw plate (left) and after compositing (right)
Figure I, shows the original raw footage compared to the final composition, figure II. The process involved match moving, colour grading, mattes (lightwrap, edge blur & multimattes), 3D set creation, texturing, dynamics and rendering.
Process
Filming
As mentioned, this was and extra shot apart of the Small Boy project. The footage is all hand held walking through an abandoned factory mill. Shot at full 1080p, 24fps on a Panasonic AG-HPX171E.
Match Move
The pillars in the foreground and background provided a lot of parallax so the footage was simple to track. Pixel Farms PFTrack was used to recreate the camera digitally and then exported to Maya. I converted the HD footage into a low quality proxy JPEG sequence. This proxy sequence was edited with increased contrast and brightness to improve the track accuracy. This is because high contrast between pixel values are good points for the tracker to cling onto.
Figure III: PFTrack screenshot after solving the camera
Digital Set Recreation
After filming I measured the set in detail (too many measurements is better than too little). I used this information for reference while recreating the set. The set did not need to be inch perfect. After all, its purpose is to cause realistic reflections and shadows to and from the spheres. As the light was quite diffused, casting soft shadows, there was a larger margin of error.
The PFTrack camera was imported and locked (to prevent accidental movement). Primary shapes were used to mock out the set, shown below. The PFTrack camera projected the footage onto the geometry (as surface shaders that don’t react with light). A reflection rig was set up around the set (ceiling and walls out of shot). After all, this is what the spheres will reflect. One large area light was positioned across the windows. Using 1 large light source rather than 8 smaller ones reduced render time but achieved the same effect.
Figure IV: Modelled digital set with tracked camera
3D Modelling & Dynamics
Modelling spheres is quite the challenge, but I pulled it off. They were placed into different z-depth positions within the digital room to give the illusion of depth with parallax. Toward the end of the scene I wanted the planets to fall to the floor, like gravity had just been switched off. Within Maya a ‘gravity field’ was applied to the spheres and a set to activate at frame 120. To prevent them from falling through the floor a ‘uniform field’ was applied to the plane. I set the amount of bounce according to the size of the sphere, so smaller light planets bounced more.
Texturing
The textures were provided from Planet Pixel Emporium. I applied them to the diffuse of mental ray shaders (Mia_X_Passes). Because Maya operates with a linear flow, the images had a gamma correction applied (0.455) to convert them from monitor space to linear. The planets were textured to appear like big floating marbles rather than real atmospheric planets. This was just to make them something different and quirky rather than serious.
Rendering
The sequence was rendered as linear images and then gamma corrected (2.2) to monitor space to match the footage in post. Understanding the linear flow allowed me to produce CG images much more accurately lit and have greater control. Mental Ray rendered in passes (diffuse, specular, reflection and so on) at 32 bit depth. A multimatte was also produced with each channel applied to planets and moons of different sizes.
Figure V: RGB multimatte with each channel applied according to scale
Compositing
The passes were added together and then multiplied by the matte. The matte being the multimatte split into separate black and white bitmap masks and then added together. This left me with the final render with transparent background, which was merged as the foreground onto the background footage. I always like to get my renders to be over the top, being too reflection for example, and then bring it down in post. It gives more control without rendered again.
Figure VI: Rendered passes (diffuse, indirect, reflection & specular)
Something is worth noting because I had an problem with this. When imported into Fusion (at least up to version 5.2) EXRs have their red and blue channels swapped, which is most problematic at first glance. This can be quickly changed in the loader settings, but is annoying when the cause is unknown. This may not be an issue with newer versions.
Because 2 posts move on top of planets a matte was needed, and this was done with a rotoscope.
The floor shadow was rendered as a black and white matte. This was input into a colour corrector node. Shadows not only darken but also add saturation, which was put into the colour corrector node. To check the shadows are the same colour, I sampled shadows on the footage for reference. Because in this scene the shadows are diffused and soft, I rendered them at a lower resolution and lower samples (reducing file size and render time) and applied a couple of pixels blur to it.
An edge blur was applied around the edge (of course) of the planets. The matte for this was created by subtracting the matte by a scaled down version of itself (inverting and them multiplying would also work). This gave me control over the thickness of the edge (by changing the size of the smaller matte). The sorbel node, which finds edges, doesn’t allow for this control and was producing edges too thick on the smaller planets (with the multimatte I could separate them and do a less intense blur). The blur was applied after the CG had merged with the footage. Grain was applied after that with the same matte.
Lightwrap was also applied. Again, separated into three lightwraps (small, medium and large planets). Smaller planets had smaller wraps. The matte was multiplied by a blurred (feathered) inverted version of itself. The original footage was blurred 30 pixels, or so, and applied on top of the CG with the lightwrap matte as its mask. The purpose was to get some of the real life environment to bleed into the CG.
I played with the idea of having atmosphere around the planets. I wasn’t sure on the look, but decided to have it but keep it very subtle. I used a similar method to light wrapping to get a matte that glowed around the planets and applied it as a mask to a colour corrector node.
Figure VII: Fusion nodal flow
A simple nodal flow, as shown above. Flowing left to right, as the eye reads, all snapped in place with each node named correctly (I find spending some time keeping it organised saves a lot of time and stress towards crunch time). I also like to make each loader larger than other nodes and with a thumbnail so I can quickly see what each branch of the flow is doing. So that’s everything. Leave a comment below!
- Credit
- Liam Major Filming, Dynamics, Tracking & Compositing
- Planet Pixel Emporium Textures
- Software
- Autodesk Maya 3D model, texturing, dynamics & animation
- Eyeon Fusion Compositing
- Pixel Farm PFTrack Match move
- Client
- Personal portfolio development
- Duration
- 2 weeks
WP Themes
4 Jul, 4 o′clock
Genial dispatch and this fill someone in on helped me alot in my college assignement. Thank you on your information.
Liam Major
5 Jul, 2 o′clock
No problem… I think.
Walid Benoihi
28 Jul, 11 o′clock
excellent … good vision
Liam Major
29 Jul, 9 o′clock
Thank you