2017 - 6th Place

"Musashi's Five Rings"

by andy davis - los angeles, usa

From Andy:

Background:

The Five Rings were five scrolls that Musashi Miyamoto, the 17th century swordmaster, wrote towards the end of his life to pass at the behest of his benefactor.  At this time, the glory of battle was over for Musashi, he focused on sharing his knowledge with his students.
He argues that as warriors, we practice our craft not for the sake of the next match, but instead because in doing so, we connect to the void/emptiness, a buddhist state of mind connecting to cosmic consciousness.
Musashi trained daily up until his death.  The sharpening of his craft was his clarity of mind.  He emphatically states that reading the words is not enough to understand.  We understand by doing.  This must be done with our own hands.

The book dovetailed with having recent creative clarity on VFX craft and workflows.  I’ve been drawn to proceduralism and apps like Houdini for approaching the problem solving of big data sets.  Shot tracking software can be viewed similarly, because it is the strategic gameboard of the VFX battlefield.  When we view from this vantage point, we need to solve the problems with the greatest clarity and avoiding distraction of granular details.

Notes:
My setups are very modular, with sections designed to swap out easily.  
The result is more of a proof of concept than a final sequence.  Cameras, characters, animation, textures are intended to be updateable assets that can be added near the “floor” geo by copy/paste from a lighter action setup nearby. 

Timing is played as an updateable asset as well with expression linking.
Was not able to apply and test across the board before deadline. 

Expect nodes to be sprayed seemingly randomly into the destination schematic when copy/paste.  If you group nodes in action before copy, you’ll have less nodes to collect on destination side.  Once cleaned up, attach to the root axis where the previous asset was and delete.  Obviously, make sure that the new asset is flagged renderable in the appropriate passes that matched the older asset. It would be great for Autodesk to implement updateable, versionable, digital assets similar to the way Houdini of game engines works with them.  This would be a far better solution.

The idea was to structurally design the project so it could be updated with new assets, animation, etc. right up until the deadline.  Work on a discrete element separately for greater interactivity that could be plugged into the great beast of the master setup.  Try not to do too much actual work in the hero setup, as this increases likelihood of app crash.

This setup currently takes me 2.5 hours on my punch-drunk macPro trashcan to open before it regains interactivity.  Not sure what’s going on, tho “bsdtar” is an active process in my ActivityMonitor.  Next week I’ll clean up the setups so they are easier to follow and encourage peeps to update w animation, geo and textures.  I can also include asset sub sections to make easier for people to play with.  Each section takes 2-6 hours apiece to render on the mac.  The Fire section with the armored warriors was perhaps the point when the hardware started coughing blood and was not able to deliver by deadline.  In general the Fire and Buddha sections contain the most geo and should be manipulated cautiously.  Perhaps this will be less of an issue on a Linux box.  

There is a point where Space-A and Space-F stop working.  Use the compass priority editor to frame the spaces when this happens. A cleaned up version of setup will keep the distance of the nodes closer, impractical when actually working swapping sub setups and dealing with action nodes when uncollapsed have many dozens of outputs.  It’d be interesting to see what kind of monster came from “clean up schematic” with this setup.

Includes an animation rig that allows ability to take character dependent geo as I built and updated them.  The rig is linked with lookAt expressions (appartently lookAt links in schematic will crash on copy/paste), and there are CTL axis for the rotation of body parts. Think this is more FK than IK, but worked in a pinch.  I plan to link the controls to a Variable matchbox shader, then linked to my tangent panels because analog input is great for animation and cameras. 

The characters created included armored samurai, martial arts students, a skeletal Death, an artist who picks up the book, a giant hand with a paint brush and a buddha figure.  These were duplicated with both replica nodes and old-school expression linked axises where replica broke down.

Animation was only able to be introduced in a few instances because of time constraints and the need to integrate into the master setup.  What animation made it in time seems too subtle with all that’s going on.  Look for it on the opening character, some of the students and the Death character, as well as the writing hand with sumi brush. 

I knew I had to come up with a look that was stripped down and decided upon a woodblock print type of look.  To get this I had all outputs active on my action nodes by default.  Creating muxes w multichannel groups with hidden links enabled me to connect them once and then Duplicate w Connections (Ctl+Shift+D) and never worry about them again.  Since they were always all on, I never needed to go back and turn on/off.  I could choose whatever pass needed for the “look” section, which could be treated in a sandbox-style area.

Aside from the god-awful load time of the setup, it renders pretty reliably and reasonably quickly, considering the heavy limitation of the ancient D700 gfx cards in the mac pro.  

There are a bunch of elements that did NOT get implemented by deadline:
    -linked cameras so that the z-depth pass could make for an interesting dissolve between scenes.
    -armored troops with 6 different characters in the Fire section.  Opposing armies w expression driven procedural restlessness.  Was tantalizingly close to getting the army in the fire section working w armor.  A crash caused 13 hrs. To recover from.
    -trees and foliage.  Concepts considered but time ran out.
 -additional couple of pieces of architecture.  Japanese aesthetic makes these reasonably easy to build, but time was taken by implementing anim rig.
 -animation for all characters.  Using anim rig it should be very interactive to create before plugging into the monsterous assemble.  Tangent panel controls with constraints clamped should make this fun, interactive process.
  -distinct kimono robes for students, samurai and Musashi.  
 -refinement of look.  I wasn’t able to spend the time to ponder and polish.  I encourage users to play with these.
-the weird eyes element in the water section are intended to focus at camera at a certain frame, then go back to watching an animated duel.  It’d also help to have an occluder element of brows to make the eyes less “detached”.
-Scott Balkcom particles.  Snow would be pretty easy to throw into the wind/void section.  Sword hits and blood splatters would be a great element, but prolly take some time to implement satisfactorily.


Misc notes:

 -animation window can get super heavy.  It’s helpful to use the filtering options to hide channels you aren’t working on.

 -someone at Apple needs to push eGPU options including nVidia.  Also, give us some new fucking hardware.