Although we’re officially on a short hiatus, I am still doing a little work on
weekends, and we have also recently been joined by Chris Kuhn, a very talented
mechanical modeling artist who is currently working on the Soyuz launch system which
features prominently in our pilot episode. I’ve also had a chance to do
some more tests with character models, shading, lighting, and rendering
— pictures, links, and video after the fold…
Animation and Rendering Tests
Officially, we’re taking some time off for me to finish work on
Lib-Ray, but I do still have some weekend time to spend on Lunatics, and
I have been getting a few things done. Here I’m going to share some of
the rendering and animation tests I’ve been doing, just so you can see
what’s going on. Please bear in mind that these are still very
incomplete, but we are getting close to ready to do some final animation
which we will use in our crowd-funding campaign, which will probably
start next month.
Welcome to Chris Kuhn!
I am delighted to have the help of Chris Kuhn, who is a really talented mechanical and science-fiction modeler. His
portfolio on BlendSwap is well worth your time to have a look at. He’s put a particular focus on
rigging mechanical models with constraints (rather than using armatures), and he’s now working on the models for
the Soyuz launch system, which plays a prominent role in our pilot
episode. He has already completed the Soyuz Launch Vehicle!
Shading for Characters
Early-on in this project, I decided to go with a “non-photorealistic”
look, but until now, we have not attempted to get this look in any of
our renders. Last weekend, I did some experiments with adjusting the
materials and shaders in our existing character models (Georgiana and
Hiromi) to get a good rendered appearance. I experimented with FreeStyle
and with Blender’s internal edge-rendering, but after comparing several
tests, I think the unlined versions look better, although the “toon
shading” produces an excellent soft feel for the characters, which I
really like.
Soyuz Interior Set
I am also re-integrating the set for the Soyuz Descent Module
interior set which will appear in the pilot episode. We attempted to get
this set ready last Summer, but there were significant problems with
combining the elements for it: scaling, file-linking, and rigging
problems. The result was a set that was incomplete and not very much
like the real Soyuz capsule. I’ve nearly managed to get this correct,
although there are still a few problems to work out.
(Compare the above real photo to the test-render below)
As you can see, we still have some work to do to get it right (you
may notice that the flight engineer’s couch is sticking into the wall!),
but then, that’s why we do test rendering! It is very tight inside the
real Soyuz module, so this is a very crowded set. I think that,
considering that none of us has ever actually been inside a Soyuz or a
trainer, and we’re doing all of this from reference photos and a few
plan views (which, frustratingly, do not always show the angles we want
or omit important details), that we’re not doing too badly.
Video Animation Tests: Evading the “Uncanny Valley”
An important concern for me from the beginning of this project has
been “will the character animation read well enough?” Specifically, we
have the combined requirements of using accurate proportions on our
characters to fit the very realistic settings and a very low budget for
animation. When very realistic models are animated very simply there is a
creepy aesthetic effect, commonly called the “uncanny valley”, that
we’d really like to avoid.
This effect is used intentionally in horror films to creep you out
with dolls or zombies which look realistic, but don’t move
realistically. And some computer-animated films run into this problem by
accident — it can become very distracting. We obviously don’t want
that. But how can we avoid it?
The effect tends to go away if either the character designs are less
realistic (more cartoon-like), in which case our brains accept a higher
level of interpretation and don’t expect the characters to move very
accurately, OR if the movement is accurate and lifelike enough to sell the character design.
We took a very risky path with the character designs on this project,
because they’re really quite close to real proportions, and as a
result, I’ve been quite concerned about how this is going to read when
animate. This last weekend, I finally got a chance to animate a brief
sequence of facial expressions and head movements with Andrew Pray’s
rigged “Georgiana” model in order to find out. I found that, actually,
it’s reading pretty well — surprisingly well, really. This gives me a
great deal more confidence in the feasibility of this production!
So, I’ll leave you with a brief animation test real that shows this,
and also my first attempt at animating Chris’s Soyuz rocket model (this
is a very rough first cut of the “pod staging” shot I have planned for
the pilot, although just about everything about it needs more work):
Video on YouTube
Video on Vimeo
Altogether, these last few days have gone a long way towards
convincing me that this project is really going to work. In about a
month, I’ll engage in convincing the rest of the world of that — or at
least enough of the world to fund our production. Thank you for
your continued interest and support!
