renderdigimania case studies - bradley and bee
TRANSCRIPT
They had a Director and a script; we had a 3D team and software which promised ultra fast rendering but which had, as yet, been untested in real production.
The challenge was to pop up an untested real-time production pipeline and render 11 minutes of high quality animation on a couple of mid-level Dell desktops.
Open the book and fasten your seatbelts. Seven-year-old Bradley and his five-year-old sister have a magical pop-up book which transports them into a wonderful world of adventure aboard Victor, Bradley's trusty toy model plane.
The production focused on two distinct spaces which would ultimately form a template for the series - a fantastical, in this case desert, Book World adventure was bracketed by opening and closing segments set in the characters’ Playroom.
The team was formed from a three-way collaboration between Digimania, Red Kite Animation and Super Umami. Red Kite, owners of the IP, did pre-production, audio and delivery. Super Umami and Digimania shared the 3D duties and took the Playroom and Book World respectively.
Claire Quinn Emily Pollacchi
Kevin McDade Martyn Gutteridge Rik Goddard Andy Reid
Navis Binu Pat Imrie Alan Caleb Michael Friel
Barry Sheridan Sueann Smith Martha MacDiarmid Ruth McKie
Garry Marshall Hugo Cuellar
The episode was 11 minutes long. Of that, nine and a half minutes – the Book World – was to be rendered inside RenderDigimania while the remaining one and a half were rendered with Mental Ray.
We had to be fast and flexible so the pipeline was split into two halves. Characters and foreground objects were created and rendered in RenderDigimania, backgrounds were 3D After Effects elements.
Our principal character modeller was offsite but quickly grasped our “game engine” requirements for modelling. So, the main character, Bradley, was modelled as a fairly low poly asset (to help with rigging and morph targets) which would be subdivided later. Confident that we had an asset that would easily pass down our pipelines, he was handed off for rigging and texturing.
3 meshes: hat, head, body 40,098 triangles unsmoothed 160,392 triangles smoothed 6 materials, 20 textures maps 52 morph targets 2 UV sets per mesh
Textures were painted in Photoshop or baked in Maya, and were usually 4096 square. One or two smaller textures were used in inconspicuous areas. Textures were exported to RenderDigimania’s project library and materials were then applied to the actors.
Our characters each had three to five materials applied, one for the face and the rest spread across the body.
We baked ambient occlusion and used normal maps which meant that the characters could not have overlapping UVs.
All textures were PNG or TGA Characters had diffuse, specular and normal textures
Our characters were set up with a fully modern animation rig. We had FK/IK switching, a muscle-based (morph-driven) facial system, squash and stretch, pose space deformers and even some dynamic joint chains such as Bee’s pony tails. We had to keep RenderDigimania’s game engine in mind, however, so the animation rig drove a render rig – a simple joint chain that was smooth-bound to the character geometry.
We called the rigged character an “actor” and it was published to the animation team and RenderDigimania’s project library. The actor was ready for animation.
FK/IK Switching Pose-space deformers
Constraints Muscle-based face rig
Dynamics
Joints Morph targets Weight maps
+ =
The rigged actors were handed to the animators who used Maya to animate them. We followed a traditional layout-blocking-animation process. Once we’d reviewed and polished the shots they were exported to RenderDigimania. The exporting process was automated with a script which baked animation down to the render skeleton and exported separate FBX files for each character. The script also baked and exported the camera.
The pilot was split up into 18 sequences, 16 of which were lit and rendered with RenderDigimania. To keep lighting consistent, an initial lighting rig of three directional lights was set up and saved as a favourite to be used as a starting point for all 16 sequences.
Individual shots sometimes required additional lighting. This is where RenderDigimania really ‘shone’. There was no need to do test renders after every lighting tweak - the results could be seen immediately on screen.
The Light Channel feature was used to great effect in sequences which had a lot of characters and camera cuts. Objects could be lit in isolation which gave each sequence the depth required to achieve the look of the Book World.
Because we were working with dynamic, real time lights and cameras, rendering became trivial. The process of rendering was no longer required for reviews or dailies. Rendering happened only if someone downstream wanted frames, and it was fast.
We rendered at roughly 5 frames per second, so we could tweak and iterate while the kettle boiled. Here's a comparison between the rendering speeds achieved by Mental Ray and RenderDigimania.
(RenderDigimania)
3.65 7.3 63,502 0.21
(Mental Ray)
408 1,633 7,000 210
Fast outputs for review, change and approval
process
Faster change iteration process. Reduced costs
and better deadlines
Games engine draughtmanship.
Using texture-based materials
Achieve everything with bones and morphs. Naming
of things has to be tighter
Apart from the export process not
many changes
All rendering and lighting
happens inside RenderDigimania