The Mill’s Suraj ‘Sid’ Harrington-Odedra takes us inside The Chemical Brothers’ Wide Open
We were lucky enough to grab a few minutes with Sid, co-3D Lead Artist, to talk about The Mill’s awesome work on this beautiful music video.
How did you get involved with this project?
Sid: Directors Dom&Nic from production company Outsider had worked with us on the Christmas campaign for Mars and were extremely pleased with the work from the team. So when they were approached by Virgin Records to make a video for The Chemical Brothers - “Wide Open”, being produced by John Madsen, they came to us in the very early days as they knew it was an extremely ambitious project.
What were some of the big challenges for you?
Sid: We knew that we would need to use a whole host of cutting edge tech in order to complete the project on time… so we threw everything at it. On set, we had Duncan Lees from Efficacy4D to lidar scan the set, which we later projected our HDRIs onto to light our subject. FBFX body scanned the dancer, Sonoya Mizuno, so that we immediately had a strong start at both the modelling and texturing of our hero character. We also had Ivo Slivkov from Timeaware fly in to get an initial motion capture pass of the dance routine. This was used as a base for the body tracking company, Peanut FX, to get started on providing us with a highly accurate digital version of the entire dance routine in 3D space. And we knew right from the get-go that our renderer of choice was going to be Arnold for Maya.
Why Arnold?
Sid: Being able to iterate quickly is the most important thing in VFX. When you end up having complex scenes that feature tens of millions of polygons, motion blur, depth of field, sub-surface scattering and displacement, you need to make sure that your ‘time to first bucket’, and your render times generally, are as low as possible. This is one of the main reasons why I consistently choose Arnold over other render engines.
For this project, we ended up needing a lot of the features that Arnold afforded us. This was one of the rarer moments for us when we decided to render all of our frames with depth of field, motion blur and lens distortion all in one go. This was mainly to reduce the processing load on the 2D team, but it also helped increase the overall quality of the final image. By not having to cheat the motion blur or the DOF, the final image looked better for it. There’s no way that we would consider doing this with any other render engine.
Tell us about some of the lighting techniques you used
Sid: With MtoA, we were able to project our interior set HDRIs directly onto the lidar geometry, place our character inside that space and immediately see some realistic looking results. This meant that when it came to look developing the mesh and clothing materials, we would be able to get great looking results in front of the director extremely quickly. The more recent integration of trace sets into MtoA helped a great deal, also. When having to deal with live action integration work, the need to be able to trace more complex paths, especially with 3rd party shaders, was extremely useful for dealing with artificial shadows cast onto the live action surfaces.
Due to the interior nature of the shot, using a skydome light wouldn't have been very efficient. Instead, using a directional (distant) light for the sun, then projecting the set HDRIs onto geometry and bouncing light onto our model yielded the best result.
One of the trickier technical challenges of the project was faking the light that shines through the mesh. For this, we came up with what we called a "stained glass" pass. The final mesh ended up having a piece of geometry shrink wrapped around it, which was only to be used for the purposes of outputting technical passes, so was never directly seen, but had to be rigged with the dancer. Having this 'stained glass' geometry meant that we were able to change the properties of rays when they traveled through it. We coloured the rays red when doing shadow passes, meaning we could isolate the light traveling through the mesh. We would then use this shadow pass in 2D to grade the live action skin and clothing.
