A 3D Character is born !

What does it take and what is behind most 3D animated characters we see on TV and movies this days ?

Despite me having been in the industry of 3D design and animation for a good 25 years, the final pieces of the puzzle begin to fit together only in the past few years. Character animation has never been my forte and for this I have skipped and avoided the topic for most part of my career, but maybe also out of fear and lack of knowledge !

Its complicated – as one could say ! Only until one gets to understand the key principle and concepts, that I try to shed some light on with this article.

The beginning.

A character needs to be born, grown and matured. For me the best way is with pen and pencil – or at least as digital hand sketch. 

The character needs a personality, like any character, for a screenplay, book or animation. The personality can reflect in physical traits or by the expressiveness and body-language.

In a corse I took a long time ago at PIXAR, they thought us, that 3D visual do not need to be physically accurate nor photorealistic, they need to be authentic and plausible !

A lot of the authenticity is created by plausible and authentic behavior. My little TVman character above has a old vacuum tube as torso, its very unlikely that this tube would bend like a human torso with spine would do, lending the character a certain mechanic stiffness.

To counter the stiffness the character is built around some loose fitting coils that can wobble or vibrate. A loose hanging ‘ponytail’ of cables is dangling from the back of the head, swing and bumping against its body and surroundings.

A further character element is the face, a CRT screen that is supposed to faintly glow, with a mouth resembling the bright spot on a old tube that afterglows when turned off.

3D model:

Some considerations need to be taken into account when a 3D model is created. This is where experience with character animation becomes very quickly important. Without some insights a character model can become a nightmare to handle further down the line, or won’t deliver any useful performance at all.

The character needs to have as much detail as possible, but remain as lightweight from a computational point of view as possible, else animation will become very slow and painful.

Do not start with the modeling before you don’t know what the model is supposed to do. Which movements does it have to perform well, which ones less. Which features will it require to tell the story. Buzz Lightyear needed a laser ray gun, a feature built into the model.

Going back to add features at later stage can mean to recreate the entire structure and animation from scratch !! I learned from my past mistakes. 

I knew my character needed a bit of a spunky personality, initially I sketched some trends of hair as image onto the CRT screen. Later I decided some cable strands dangling in its face could add that little extra character. He can flick it out of his face, only to slide down and cover half of it again.

The hands where born out of necessity, without much creativity behind – i am still thinking of a better set of tools, but for now the ‘claws’ have worked well.

His spunky personality makes him a bit edgy – after all he is one of kind and he knows that too well ! He constant taps his feet, he needed extra control for the feet to tap his toes.

Since the little TVman is mechanical little emotional expression can be given, for this the glow of the vacuum tube got a soul, a faint glow that pulsates like a heart beat. Different speed and color gives him the excitement needed.

Beyond that, head control – pitch, heading, roll – where needed. Arm,leg and hand controls as any biped would require.

Structure:

With the above traits in mind, I was ready to build the model. Generically a model ends up being one solid polygon mesh for the most part, so it can be easily deformed and rigged. In this case I chose a slightly different approach. I wanted to keep control over the coils and various  other mechanical parts of the body. For this reason the model remained more broken down into components.

I ended up turn the arms and legs into solid meshes, while keeping most of the torso and head as individual elements.

Nevertheless they need to be hierarchically ordered, so they can follow one another movements without falling apart. The torso became a child of the hip structure, containing its cables and coils as elements. Legs are a child of the hip, so are the arms and head of the shoulders.

Most other models I worked with where one combined mesh, which makes it much more easy to work with.

Rigging:

Rigging ! A process I learned way too late in life, but the most important step to before one can make compelling animations. Since this is a largely mechanical, non organic model, I could have left the elements of the arms and legs hierarchically structured and animated them in a traditional forward IK manner. One element follows its parents position and animation. Since this model has cables wrapped around its limbs, the movement would result in kinks or other undesirable behaviors.

I opted to bind the limbs into polygon meshes so they would later deform nicely.

Rigging refers to the task of adding a skeleton or bone and joint structure to the model. Bones are drawn separately from the model to form a skeleton like above. This bones can then be animated and the polygon mesh follows and deforms accordingly.

The limbs are straight forwards traditional biped structures, hip, knee, ankle joint, to and shoulder to wrist.

In order to add some dynamics to the hair in the face and the ponytail, a fairly detailed chain of bones where setup for each strand and cable.  I added dynamics to them so that they follow some physical properties like drag, gravity and stiffness/softness. In short when the body swishes around, the strands bounce and follow as they would in the real world. A perk of modern 3D design, we don’t have to do this behaviors manually any longer, they are generated automatically once the rig is configured the correct way.

Some invisible collider object prevents then elements to intersect with the body it self, making the hair and cable repel or bounce off.

The next step is to bind the bones to the elements of the 3D mesh. A simple task if done right. All it takes is a bit of cursing and practice. It took me few hours until this model worked to my liking.

Controls:

Controls, a topic on its own. As mentioned above, the model needs character. Knowing the elements required makes it much more easy to decide which controls will be required. Any object can be controlled manually, object by object, but that is too tedious during animation.

For this brining out user elements for the controls makes life 100x easier !

From previous jobs I knew pretty quickly what I wanted to control and in which way. The little sliders in below image show the majority of elements that I ‘wired’ out to give direct access.

Arms and legs will be controlled via motion capture data, therefore they would not be required for manual controls. I will most likely end up removing the leg controls, but add the manual arm controls, as this will be done to a lesser extend via MOCAP.

Hand controls will all be animated manually, for this  I only needed open/close and rotation, since its simple claws versus humanoid hands.

Invisible in this image are controls that make the head look at the camera automatically if I desire, a simple switch in a XPRESSO control panel needs to be set to enable it.

Animation:

As mentioned above, a lot of the animation will be done with motion capture data. I own a a small motion capture system, based on a Microsoft Kinect sensor. Allowing me to jump and run around while capturing the data of the most important joints for arms and legs.

In order to map the incoming data to the model I needed some translation of the vectors. Mainly range mapping to translate the different scales of my human proportion to the model. This was done with a XPRESSO node system in Cinema4d. A simple node structure that can be switched from different source of data.

This sources are in my case FBX files from the Kinect, to other resources, each one might have a bit of a difference scale, hence many ranger mappers are present and some monitoring nodes to make trouble shooting easier.

This structure took me few hours to setup, but will safe me hours up on hours during animation !

It looks convoluted but its actually fairly simple ! I designed much more elaborate ones, mapping phnomens, mouth positions per letters to facial expressions. Where I would only have to enter a phnomen like “th” and it would morph the facial expression to the according position. A system like that can take days to setup, I might end up copying it onto this rig to ease animation of the mouse and eyes.

This is it ! 

From this point onwards its plain and simple animation. If one figures out the above steps, each software has different approaches and mechanisms, but the basic steps are the same.

I work with Cinema4D since 25 years. All of the above can be done under one roof and many many more.

The software alone might be daunting at first, but overall its not extremely complicate with a bit of right insight. Which I hope to have provided to a certain extent with this article.

Feel free to ask me questions. I wish I had somebody to ask, all of the above I had to figure out over many years on my own. Therefore, there might be better ways of doing things, but the above has worked for me for many projects on a daily basis.