Nyeng!

…met with a massive headache.

Roughly matching the angle of a real camera?  Easy, thanks again to the towel.  Resculpting Daisy’s breasts to help achieve a pose that her rig isn’t actually capable of?  Also easy (easier than matching the camera, in fact).  Rendering without any problems?  …crap.  I tried to simplify my approach to rendering with this one, but only succeeded in making things more difficult…

So far my re-write for Mary-Anne includes two things she’d been lacking - a job, and a hobby.  That was one of the painfully obvious problems with her, before - she didn’t really have anything to do.  Now I’m picturing her as a carpenter/handy-girl, so she might as well dress the part.

This costume change was originally just going to be for the winter-themed short I’ve been planning, but now might be a little more permanent, with a few changes.  Instead of long-johns, she’d have a t-shirt, and she’d also have heavy work boots instead of homemade squirrel-shoes.

In Maya, the costume change might also make things easier.  With more of her body covered up, I won’t have to worry as much about corrective shapes.  Baggy clothes, if I’m using nCloth, also means that I don’t have to worry about getting the skin weights just right, to prevent the clothes from intersecting her skin.  And the boots will make my life easier, just for the fact that I won’t have to worry about her toes any more.

How do you write a character that you just cannot understand, or even answer basic questions about?   That’s the problem I’ve been having with Mary, for quite a while now.  I haven’t been able to use her effectively, because I don’t know anything about her, even after working on her for a few years.

All of my other characters, even more recent ones that I haven’t drawn yet, that only exist as notes on an index card, are easier to write.  One thing they all have that Mary doesn’t - real personality traits.  Each of them represents at least one obscure part of my personality - Daisy is my determination to overcome any challenge.  Bill is my irresponsibility.  Mercy is a combination of my apathy, and barely-subdued anger.

With Mary, I can’t think of one thing that she’s supposed to represent.  There’s no basis in reality, no emotional connection, nothing that really lets me get inside her head and know exactly how she thinks.   She’s a prop.  What’s her favorite color?  Favorite food?  Life goal?  Worst fear?  Hell if I know, but I’m working on finding out, by starting from scratch.  I’m throwing out her current backstory, and writing a new one that will hopefully not only make a lot more sense, but also make it easier to place her in comics or a short film.

Fortunately, there’s a silver lining to how poorly she was written up to this point - she fit in with the other characters so badly, that rewriting her doesn’t affect them much, if at all.

Of course, there is also a way that I could keep her as-is, where she would not only make some sense, but would also allow for some twisted humor - what if she was actually Daisy’s imaginary friend?  It could make sense in that she has the type of body that Daisy wishes she had, and is also completely care-free, with none of Daisy’s responsibilities or troubles.

There would still be a couple problems, though - if Mary is imaginary, then she can’t be in any stand-alone stories, and other characters can’t interact with her.  Another thing is the fact that Daisy is weird and interesting enough already, without schizophrenia being added to the mix.  And then there’s the fact that this type of thing has been done before - Calvin & Hobbes, Fight Club, Ferris Bueller’s Day Off (according to some people), among others…

One of many things I’ve been meaning to draw for a while, to help develop my characters a little.  Bill’s a late addition, though.  The original plan was to have Daisy continue on a little longer, until Mary interrupts her with a fit of brain-freeze.  Bill seemed like a better fit, even though now Mary is little more than a prop.

A recent discussion on CGTalk has made me take a second look at how I render my scenes.  For the past couple years or so I’ve preferred to render out multiple passes/buffers, and combine them in post.  First it was with ctrl_Buffers, and lately it’s been p_MegaTK.  Now I’mwondering if it’d be better to just do everything in-camera.

I guess you could say this started when Final Gather and MegaTK caused Maya to crash.  I’ve also been a little frustrated with MegaTK’s limited light support.   Through its own light shaders, MegaTK’s buffer output supports Direct, Point, and Spot lights.  Image-based lighting and mia_physicalSky can be used for ambient/indirect light, bit not as a primary light source.

Another possible reason for my reliance on passes, could be that until recently I’d been using the wrong settings for subsurface scattering, and also because I didn’t know about MR’s linear workflow.  While I’m at it, I’m tempted to give V-ray a look as well.  Just as soon as a demo is made available for Maya 2011 (and hopefully before my 2011 demo expires)…

Not long ago I started playing with the demo of maya 2010.  Now, it’s 2011’s turn.  With 2010, I was mostly focused on seeing just how far I could push the 64-bit renderer.  I also played around with nCloth and parts of the muscle system, but that was about it.  With 2011, I’m trying to be more thorough.

One thing I was curious about was Assets.  It’s one thing to read the description, but I had to see it in action to really get what the big deal is.  At first glance, they seem to work just like group nodes, if you’re just adding objects to it.  Publishing an attribute from one of those objects to the asset node seems like it could be done with a script in older versions.  Where the asset node really shines, is in the Hypershade.  I created several shaders and utilities, added them to the asset node, and POOF!  Gone!   Double-clicking the node opens it up so you can work with the nodes.  Now I understand how this could benefit characters - rather than having the hypershade littered with dozens or hundreds of utility nodes, all those nodes can be tucked away inside an asset node, drastically clearing things up.

Animation layers are something I’m almost ashamed I haven’t tried sooner.  They were a regular part of my workflow in Max, to the point that the may have become a crutch.  Whenever I tried animating in Maya, I missed that feature.  From what I’ve seen so far, Maya’s animation layers work very similar to Biped’s in the ways that matter, and add in a few very useful tricks.

The new skinning tools also deserved some attention.  If I can finish cleaning up Bill’s model, I may have to use him to fully test out the interactive binding tool.  One thing I was able to try out now, was the dual-quaternion skinning.  Since it’s built right into the smooth-skin node, this meant that all I had to do was load one of my characters, and switch the method from linear to dual.  Here’s a comparison of the two methods on Daisy -

While it’s far from being a perfect solution, the dual-q mode at least seems to help maintain volume.  The verts will still need to be carefully weighted, the only difference now is you may be able to get away with weights that would look terrible in linear mode.  Pose deformers will still be needed in key areas, and fortunately that plug-in works just fine with dual-q.  I’m not sure why I thought that it wouldn’t…

On and off, I’ve been looking for a way to render soap suds for a least a couple years.  Now, I might finally be closer to figuring it out.

So far, the setup is pretty simple - blobby particles emitted from a sphere.  The particles are shaded with MegaTK, while the sphere uses an SSS shader.  The next step might be to sculpt some more sudsy-looking shapes to emit the particles from, and see just how many particles can be rendered before Maya crashes.

At this point I should be used to discovering things that will save me time in the long run, but cause a lot of immediate trouble.  It’s not any less frustrating when it happens, though.

When I started modeling that bedroom and bathroom, I it would have to be once big piece.  This was because I thought that any seams would be visible when occlusion was rendered.  What I found out this morning, was that occlusion doesn’t work that way.  Occlusion cares about angle and proximity, but if two flat surfaces are perfectly aligned, the seam won’t show.

Since occlusion won’t be calculated along the seams, this meant that the rooms could have been modeled separately, with cleaner geometry.  Things like the archways can be simplified even further, by using nurbs surfaces instead of polygons.  I just wish I’d figured this all out BEFORE I’d modeled the rooms…

Something else I can’t figure out, is a problem with the mia_roundcorners node -

It seems that if a surface is folded one way, the edges will be rounded as expected.  But if it’s folded the other way, nothing happens.  The outside of a cube gets smoothed properly, but the inside does not, even if the normals are reversed.  Unless I’m doing something wrong, this could be a problem for interior renders.   Of course, one possibility could be the fact that the roundcorners node isn’t technically supported by Maya 7…

When it comes to pinups, I have the same problem in both 2D and 3D - backgrounds.  In 2D, it could be a simple matter of not having enough practice.  In 3D, the problem is that everything needs to be built and textured.  And possibly that I haven’t had enough practice building environments.

The first environment I’ve decided to tackle may also be the easiest - Daisy’s master bedroom and bathroom, modeled after those of my parents’ house.  The blueprints are a huge help, and I can always pay a visit if I need up-close reference.  Probably the most difficult part is trying to model everything somewhat accurately, then having to squash the height by 35%.  It’s difficult because many objects/details need to be moved down, rather than scaled.

Why 35%?  Because at 3′6″, by my math Daisy is about 65% as tall as an average woman.  And for the story I’m cooking up, her house is built specifically for her, sort of like her own fortress of solitude.

One major concern whether or not I’m modeling it correctly.  I’d like to keep it as simple as possible, and avoid having to smooth it if I can.  Mia_roundcorners should keep me from having to bevel the edges, but so far I I haven’t been able to get it to work, at least with this geometry.  And with the hallway arches, smoothing might be unavoidable.

The bedroom is less of a concern than the bathroom, at the moment.  I’ve been thinking about revisiting my never-finished “bath time” images, by adding Daisy to the mix.  Mary washes herself in a wooden half-barrel tub out in the woods, but Daisy prefers to have walls.

Of course, some things I still have yet to figure out are the water and bubbles.  The water in the tubs can be faked with a simple surface and a refractive shader.  “Wet” spec maps may need to be painted for both characters, possibly for each pose.  The real challenge is the soap bubbles.  I’ve tried instanced geometry with both particles and Shave.  With Shave, the shader doesn’t calculate properly.  The shader works a little better when using particles, but even a simple scene (particles emitted from a sphere) takes forever to render.  Unless I try to go the paint-over route, I might have to wait until I have access to nParticles.

A thought occurred to me earlier, which usually leads to trouble.  For some reason I got thinking about the fast skin SSS shader, and how little I understood it.  I had first started using it with Mary, and have used the same settings ever since, which is never a good idea.

My initial curiosity was simply if I was using the correct scale conversion.  All this time, I had it set at 15, and left all the other settings at their defaults.  This at least appeared to give the correct amount of back-scattering in the ears and other areas, which I felt was good enough.  I have no idea how I arrived at 15, though.  For all I know, it might have been guess-work.  Whatever the case, the fact that Mary’s height had never been properly figured out meant that this scale factor was likely incorrect.  But since I know exactly how tall Daisy is supposed to be, it should be a little easier to determine.

As I understand it, the skin shader by default assumes that your model is built to real-world units.  My characters are not.  I have Maya’s units set to the default, which is centimeters.  Daisy stands 19.5 units/cm in Maya, but is actually supposed to be 42 inches.  By my math, the conversion factor should be about 5.45, not 15.

The depth and radius values for the scattering layers were a little trickier.  Fortunately there’s CGTalk, where a quick search offered this advice - Set the epidermal layer’s radius to half the width of a finger, and the subdermal and back-scattering radius/depth to half the length of your thumb.

For a quick comparison, here’s the old shader, versus the new one -