Competing in the music industry is tough work. Quite often an amazing promo coupled with a great track can propel the song/promo package into the limelight. Radiohead’s fantastic new release together with staggering CG, completed by The Mill for their music video, is such a combination, destined for chart–topping success.
The video, Directed by Alex Rutterford of Black Dog/RSA, features a virtually generated Thom Yorke performing the bands latest release ’Go to Sleep’.
The fully animated clip is set in a fictional, regency style, town square. The promo opens on a red flower, the only colour used in this largely monochromatic film. The flower moves gently to the opening bars of the track as Yorke sits on a park bench in the middle of the square. Office workers storm past going about their mundane routines unaware of Yorke sitting on the bench.
As the track builds up, the location begins to dramatically transform. One by one the buildings surrounding the square begin to self–destruct and turn to rubble. Oblivious to this, the crowds of people continue to go about their daily grind, unaffected by the falling masonry.
As Yorke’s performance builds, the strewn rubble of the fallen buildings begins to reform itself into an almost Bauhaus style of building – flat roofs, smooth facades and cubic shapes. The historic and opulent regency style facade is replaced with flat faced, concrete, modern architecture. Throughout this transition people walk straight past buildings without noticing a change, their route un–deterred, their focus directly on the path ahead. The promo ends as it began with the camera pulling back to reveal the red flower.
The style of the film is photo realistic in movement combined with stylised polygonal faceted textures in look. Yorke is therefore fully realistic in his performance, while at the same time being a stylised version of himself. The buildings have no real texture, yet the dynamics of their collapse mimic reality and an actual demolition.
Around twenty Mill professionals, both 3D and 2D, worked for approximately eight weeks to build and animate the final promo. The process began with Director Alex Rutterford’s lo–fi CG animatic, which entailed the video. The Mill team then used this as a guide to build on, updating the animatic with the latest scenes as they developed. Along side this, some of The Mill CG team set about modeling the various buildings, both old and new. In the mean time, others were responsible for the animated characters that were to inhabit the environment. These were split into two areas: generating the crowd characters; and of course, Yorke himself. Finally, all the CG elements were combined, lit and rendered before various layers of CG were composited in flame to create the finished film.
Yorke’s character was one of the trickiest and technically difficult areas of the promo to complete. Once his stylistic look was established, Thom Yorke himself was required for a very technical shoot and scanning day. Firstly Yorke’s head was scanned in several poses to generate a very accurate CG model of his head. His movements and performance were then captured using motion capture. The first part of the process recorded his body movements, the second, concentrated on his face, with Yorke wearing around seventy markers on his face alone. This raw material was then combined and finessed using actual video performance of Yorke and Radiohead, to create the virtual but realistic performance seen by Yorke in the promo.
To generate complex crowd scenes, The Mill’s team were able to use new beta software ’Massive’ – originally designed for the large crowd scenes in Lord of the Rings – commercially for the first time. Massive allows the animator to generate crowds of people that have virtual interactive intelligence. Each character was initially animated using motion captured walk cycles and movements. Massive then gives the individuals their own little brains to detect such things as terrain, so in this case the curbs and pavements. It also allows them to detect other individuals, so allowing them to realistically avoid each other. This means complicated interacting crowd scenes of infinite sizes can be created. While several of the periphery characters in the crowd scenes were hand animated, Massive was used to generate the large groups of people needed to fill the square.
Once the modeling of the buildings were complete, Mill animators set about creating the dynamics to allow them to be both destroyed then rebuilt in a different form. Rutterford was keen to get a very realistic feel for the buildings that were to break and crumble. Therefore, the Mill CG team used reference material from real demolitions and destructions as a guide to their animations. They then created the deconstructions using a combination of hand animation and computer dynamic simulations combined with layered particle animation effects.
In lighting the final piece, the team used its in–house written light dome shader, to give the film a very ambient realistic lighting feel. This enabled them to create the realistic shadowing and light fall–off seen in the film. Extra details were also added to Yorke’s face in several shots to create the low poly look created by a mesh reduction technique. This was achieved using a plug–in developed by Mill 3D’s R&D team.
The finished rendered animation passes were then combined in Flame. Live action smoke and dust was composited onto the buildings to enhance the illusion of them collapsing and rebuilding. The flame team added drama to the darker scenes by adding light emissions to the streetlights, along with smoothing out the transitions from solid buildings to cracked ones. They also finessed areas such as depth of field and adding of subtle motion blur to Yorke’s movement to enhance his performance.
The final touch that Mill Flame operators worked on was creating the camera shake effect as the buildings collapse. This significantly helped to increase the scale of the impact and weight of the CG in these scenes.
The finished film is one of high audiovisual impact. If Yorke’s realness both in form and performance doesn’t absorb the viewer then the dramatic changes occurring in the background bound to stir viewers’ emotions.
