Actor's Studio, lesson 5 - Pallas walk cycles

some of the walk styles

The best reference manual for walks is probably Richard Williams' book, The Animator's Survival Kit, which devotes 110 pages to the subject. Books by Preston Blair and Eadweard Muybridge are also useful.

Almost no video reference was used for these walks, so some of them look a little mechanical. However, the goal of this lesson was to explore various walks, not to perfect them. Video reference can speed up animation production, and make the movements seem more natural. For very cartoony characters, video reference isn't so essential.

Pallas' face hasn't been rigged yet, so she can't blink or emote through facial expression. In these early lessons, all emotion is expressed by the rest of the body, through posture, pace, amplitude, etc.

walk cycles

NORMAL WALK - This is just a stroll, not determined to go anywhere quickly.

In the video, I included a series that demonstrates how intermediate poses are added to build a smooth walk. This was the most difficult walk to do. For most of the subsequent walks, I just copied this walk to a new animation layer and modified it.

EXAGGERATED WALK - For this walk, everything steps, bends, and swings further. This was for studying arcs, limb twists, weight, and follow-through.

DOUBLE-BOUNCE WALK - This is a cartoonish sort of walk that doesn't exist in nature. It was popular in 1930s cartoons, and thereafter was only used ironically. I tried walking like this and it's exhausting.

SNEAK - This was one of the more fun and easy walks to do. Everything was pretty easy except for the hands, which don't seem completely convincing. There is a ton of drawn reference for this type of walk in Animator's Survival Kit, but I would probably want live-action reference if I was going to make this for a client.

PLOD - It seems I had the same problem with this walk. The arm/hand weight and movement don't look correct. Good video reference would have fixed that quickly. maybe next time

SASHAY - This was a fun and fairly easy walk to do, probably because there is a ton of video reference in the form of fashion shows.

SWAGGER - This is probably just the exaggerated walk, but more aggressive. While making it, I was mostly imagining Popeye walking into a fight. Again, the weight and follow-through aren't exactly right, so video reference would've been helpful.

RUN - The run was pretty easy to do. I mostly used Muybridge photos as reference. I'd like to do more expressive runs someday. This one is a standard fast run.

SKIP - This one was a little difficult and - again - I wasn't sure what to do with the arms.

In conclusion, I learned much. There's more to learn about walks, but I'm moving ahead to study other actions and behaviors.

next: face rigging

Actor's Studio, lesson 4 - Pallas body rig

Pallas humanIK rig

Pallas' next step was rigging.

Rigging is the process of attaching deformers to an otherwise rigid model. CGI character deformers are simply the digital counterpart of a puppet's armature: a marrionette's wood rig, a stop-motion character's metal rig, a Muppet operator's hand. Without a rig, these digital and cloth characters are inert.

I'm using Autodesk's humanIK auto-rig. One of its advantages is that it can easily be plugged into motion capture (mocap) data. That is, a real actor's movements are captured on a computer and transferred to an animated character. Mocap replaces an animation technique called rotoscoping, which was first utilized a hundred years ago and is only rarely used today.

For the video below, I plugged some stock mocap data into my Pallas character to see how she'd handle it. These quick animation tests are a good opportunity to detect problems with the model, texture colors, bone placement, and skinning. This video isn't a cinematic-quality rendering, just a quick screen-capture of the modeling space.

to watch in high-definition, view on Vimeo

These mocap files only hold limb, torso, and head animation. If an animator wants the fingers and face to move, he has to add that separately. Consequently, Pallas' face and fingers are stiff throughout this test video. I didn't spend any time correcting the raw mocap data, so - if you look closely - you'll see there are some strange positions throughout: legs penetrating torsos, arms penetrating each other, clavicles not rising with the humerus, etc.

This isn't a project about practicing dynamics, so there will be no dynamic clothing or dynamic hair on this character, or not at this early stage.

next: keyframe animation

Playground, chapter 9 - children edits

children, old & new versions

In the original version of the kids, their hands were mitts with conjoined fingers. This was partly an economic decision and partly a design decision. Conjoined fingers are easier to rig and might look cute if modeled well. However, the hand is the most expressive part of the body, next to the face. If fingers can't move, characterization becomes limited, so I gave the kids new hands with adjustable fingers.

new kid rigs

The previous FK skeletal system has been replaced with a new IK system. With FK (forward kinematics), each bone must be rotated separately in order to move a character into position. With IK (inverse kinematics), moving one bone will influence the position of neighboring bones. This makes it much easier to pose characters, so most of these new poses are a little more nuanced.

Besides changes to the rigs and poses, there were also the usual adjustments to children colors and textures.

Next in this project, we'll be adding or fixing various foreground elements.

Actor's Studio, lesson 3 - Pallas texturing

Pallas head model (top), shader (bottom), texturing (right)

Textures can evoke tactility, temperature, and odor. A geometric mesh becomes something more relatable.

These textures were created by projecting high-resolution photographs - of skin, hair, and fabric - onto the geometry. This is just a first pass. The textures will continue to be developed, depending on requirements of design, animation, and rendering. For now, they're good enough to continue onto rigging. Again, this is intended to be a simplified model of medium detail, so I avoided painting pores into the bump and specular maps, and other meticulous details.

textures for Pal's head: diffuse, bump, specular, roughness

Pal's head uses a roughness map. Dark areas of a roughness map represent a smoother part of the surface, with sharp reflections and highlights. Bright areas of the roughness map represent abrasive or porous parts of the surface, with blurred reflections and diffused highlights. Some rendering systems use "glossiness" instead of roughness, which is simply the inverse of roughness. IE, you'd need to invert your roughness maps to use them as glossy maps.

Usually, I create a roughness map by simply inverting an instance of the specular image file. For most surfaces, this is probably an acceptable technique. However, it's not specific enough for human faces, which audiences instinctively scrutinize more carefully than other surfaces.

Pallas' hand

Additional textures will be needed, somewhere down the line, for rigging skin and clothing animation, including wrinkling and blushing/pallidity. These textures will be layered images, rigged in such a way that their opacity is increased when a surface is wrinkling or changing color.

next: Pallas body rig

Actor's Studio, lesson 2 - Pallas modeling

Pallas model

Here is the finished model of Pallas. The purpose of this project is to augment my rigging and animation skill, but it's also an opportunity to become more familiar with facial anatomy, and pick-up the latest modeling and texturing techniques.

This is a character of medium complexity: halfway between a smiley face and a photographic face. Key facial details can be included to make a character more relatable. For example, when you add eyebrows to a smiley face, the range of possible expression is increased. The face that previously could only express happiness or indifference, can now feel anger, sorrow, mischief, relief. Below is a list of key skin folds that can assist in maximizing character depth.

head-modeling details

procurus & depressor supercilii - These are muscles between the eyes that are apparently only used for conveying aggression through the furrowing of the brow. In this area, the polygonal edges follow the direction of these muscles, so that the furrowing will be smooth. (For this project, the plan is to use driven displacement maps to create brow furrows, if that level of detail is required. However, that could change.)

palpebral sulcus - This is a furrow above the eye that holds extra skin, needed to close the eye. The furrow disappears when either the eye closes or the brow is raised. On Asian eyes, this sulcus is hidden by the epicanthic fold. In modeling, the sulcus can easily be created by tucking in one of the edge loops. When the eyes are in a half-open state, the presence of the sulci can help make the character look more relaxed / less alert.

lateral canthus - This is a groove on the outside of the eye where the upper and lower eyelids meet. It's minimized when the eye is opened wide.

nasolabial sulcus - This is a furrow that separates the nose and lips from the cheeks. Its the only skin fold represented in the aforementioned smiley face example (at the corners of its mouth), and beneficial for highlighting smiles and snarls.

oral commissure - This is the corner of the mouth where the top and bottom lips meet. In modeling, it's necessary to tuck the edge loops of the lips into these corners. Otherwise, the mouth looks strange and inhuman.

next: Pallas texturing

Playground, chapter 8 - second version

children, before/after

Nine months after completing the Playground, Anna and I agreed to have another go at this rendering, to create a more interesting version and solve all the problems we have with its quality.

First, we created new hair textures and a new skin shader for all the children, which makes them look less like plastic. I'm using overcast lighting this time, primarily so the children's faces are easier to distinguish.

details from design sketch

Here are some of the other changes we're planning:

• add more foreground items: animals, clothing accessories, plants
• articulate the children's fingers, so their hands can be more expressive
• renovate the barn so it contributes more to the composition
• model new distant hills with details like buildings, vehicles, and animals
• create a larger variety of trees
• paint a new cloudy sky to match the new lighting

Anna's Doll

Anna's doll

Anna designed and modeled the doll in this image. I modeled the environment and built the rendering, including lighting, textures and materials.

Due to the distressed nature of the architectural surfaces, my attention to detail needed to be more acute than usual. For such a task, photographic reference is useful as a reminder of how things decay.
In my studies, I noted that ...

• More dirt will find its way into the concave edges of surfaces.
• The convex edges of a surface are more worn from contact, making them smoother and shinier.
• More staining and damage occurs near the bottom of a wall.
• With age, the gaps between wall and floor expand and become jagged. Sometimes these gaps are filled with sealant, usually in a sloppy manner.

It was an interesting though tedious study. I wasn't super meticulous with the detail. I just tried to create enough so that the fakery isn't too obvious.

composite, before & after

For a multipass composite, a 2D rendering from a 3D program is separated into multiple images, then recombined in a 2D program. This provides greater control over the image quality, and results in better volume and spatial depth, and a more precise implementation of simulated lens effects: i.e, more realism.

render passes

 This is my first complex composite in Nuke. My previous Nuke composites just used a motion vector pass and reflection pass. This doll composite uses 10 passes: direct light, ambient light, specular, reflection, refraction, shadows, occlusion, subsurface scattering, z-depth, and masks.

Nuke node graph for doll composite

Actor's Studio, lesson 1 - Pallas design

Pallas, modeling guides

For my next project, I'm developing my character animation skills and rigging strategies. The subject for this project is Pallas, a girl of around 10 years age with normal proportions: e.g., no gigantic cartoon eyes.

A character animator is an actor who applies his performance to a proxy, similar to what a puppeteer might do; hence the project name, Actor's Studio.

Pallas' appearance will be fairly generic so that she doesn't assume a particular personality type (e.g., debutante, goth girl, etc.). For now, her hair is covered by a hat because I want to focus on simulating human behavior and not be sidetracked by hair physics.

The above image shows some of the guides used within the modeling program. There are guidelines for the clothing and skin exteriors, edge flow, and bone and joint locations. The rib cage and pelvis are noted to avoid torso deformations in those areas.

next: Pallas modeling

Ring of Fire

This is fan art for Ring of Fire, a talk show about corporate crime and legalized corruption. It's simply an animated logo with a generic camera path. The animation transitions to a blue placeholder, which is where the talking heads would go.

I used Maya fluid effects (FE) for the fire. I wanted motion blur because it adds another layer of realism. Without it, animation can look choppy and flat. FE are invulnerable to motion blur, so I had to go through a tricky process to achieve it. I'm reporting the process here because I can't find any tutorials that cover it, and maybe this post will benefit someone. (If this helped you, please say "hi" in the comments.)

There's probably a better solution, but this is the simplest one I could devise. It involves creating a mesh approximation of the FE particles from which you can extract motion data to apply within a compositing program (Nuke, Fusion, AE, etc).

MOTION BLUR FOR MAYA FLUID EFFECT

step 1: Create your fluid object. Ideally, create a fluid nCache for it.

step 1

step 2: Create a polygonal cube, or any other polygonal object. It doesn't matter which. Set the cube's transform vectors to 0,0,0; 0,0,0; 1,1,1

Wonder Warthog, part 12 - new head

(I'm redesigning, modeling, and rigging Gilbert Shelton's classic cartoon character, "Wonder Wart-Hog.")

NEW head

I did a little research of software and anatomy, and decided to re-engineer the warthog's head.
Some of the topology was improved for more anatomical realism and better deformations (ie, the ease of assuming various facial expressions).
The eyes were raised about a half inch.
I switched to more appropriate rendering software.
Alternate hair software has been employed.

OLD head

Above is the old head, which I last worked on over a year ago.
Geometrically, they are basically the same head, but a few tweaks and some new software resulted in a substantially different look.