Noodles and MorePosted in Production by andy
50 shots to go, don’t panic.
As the final deadline is getting closer and closer (more exciting, riveting, mind-warping news about that soon!) everything seems to be moving at light speed. almost the entire movie is animated and ready for further reviews and improvements, the environments are completed, final scenes set up, lit and composited to shape what’s going to be the first Open Movie of the Blender Institute: Big Buck Bunny!
*ehem* now that the PR-talk part of my brain is satisfied…
An essential part of the “finals”-process is the blender node compositor (aka. “noodle” – editing). A couple of weeks ago Nathan kissed the animation team goodbye and was transformed into a node-wrangler wizard. He is currently helping me to take the files from the animation- and environment department apart and assemble them back in compositing.
There are multiple reasons for utilizing a compositing pipeline that is integrated tightly into the rendering process:
- many scenes are way to complex to be rendered in one single pass (trees, leaves, bushes, grass, fur, feathers, huge textures, matte paintings)
- a rather annoying limitation of blender is that it does not handle motion blur and depth of field inside the rendering process very well, they have to be done in post. luckily blender has one of the most powerful 2D vector motion blur effects available, now with a new option to make curved motion blur (arcs are an essential tool in character animation)
- characters and scene elements often require individual color and effect treatments that go beyond lighting and exposure adjustments. after all, this is a cartoon!
- the anti-aliasing process is done after post-processing on ALL the rendering samples for extra crisp image quality (the new FSA option), also useful in masking and layer-seperation.
Here is a small step-by-step insight into this process:
the first step (often done parallel to lighting) is to seperate the individual elements of a shot into seperate RenderLayers: characters, foreground, background, etc.
the initial compositing setup is very basic, all elements are simply layered on top of each other using AlphaOver nodes. this can sometimes be very tricky if complicated masking / blending is involved.
effects like depth of field and motion blur are added after that.
after that we can easily import our pre-made nodeTrees, these are very specific for each element / scene and often require further adjustments. this speeds up the compositing process quite significantly.
most of the time, the situation requires color correction to be done before (motion) blur, that benefit from the full high dynamic color range.
that’s an example for a character-specific nodeTree group: a separate ambient occlusion pass is multiplied on top of the rendered image. both inputs are color corrected individually for maximum artistic freedom.
each render layer can have its own set of sub groups.
and that’s the final composite! a very simple shot like this can be composed together in a matter of minutes (rendering, setting up lights and tweaking stuff takes extra time of course). now this shot is ready for the render farm, NEXT!
A small but quite helpful detail for those who are familiar with the blender compo system: All the compositing setups are done in another blender scene, separate from the visual elements. This makes it very easy to append a setup from another shot for example. Different shot elements can also be separated into more than one blender scenes for better memory management. This way, memory can be flushed after each rendering pass to have more for the next. (activate “Save Buffers” and “Free Tex Images” in the Scene Buttons)
Since you can fill entire books with all the compo stuff (which will probably happen sooner or later), I can show you a small selection of our rather strange node setups instead:
back to work!