That’s more than can be said for my human companions. Our drunk driver Roger didn’t talk much, though he was always friendly, and the ice-queen with whom I shared a tent was 100% business. I even hung a picture of her on the wall and had the car re-painted in her favourite beige, but all to no effect. She wouldn’t even look at me…

Click here to view the embedded video.

All body animation appears to be rotation-only, given that the limb lengths do not warp and stretch to assume the positions of the bones in the animations. IK does however appear to be solved for each limb as the arms and legs of the shorter characters hyper-extend (stretch out) to meet the required feet and hand positions of the original animations. The only visual artifacts like this occur in the face, showing that position keys must be involved in creating the facial animations.

The camera animations for intros and special moves are bundled with the animation sets, highlighting that both the sound and visual effects of the chosen character remain intact on the character rather than the animation sets and play out on corresponding actions quite well. This illustrates that there must be identically organised sets of each for every character.

[via GiantBomb]

Looking at the image below, they at least animate using Lightwave, though how much of their pipeline relies on it is unknown. Perhaps the Japanese industry is different as a whole, as in the West we mostly use either Maya or 3D Studio Max. Lightwave is more favoured among high-end artists for it’s renderer, not its animation system.

The rig below includes the wireframe mesh of the shadow-casting model, but otherwise displays fairly standard IK/FK limbs, with an aim constraint to animate the head, (presumably also able to be turned off). This aim-target is of interest if it is actually exported with the animation and used in the real-time target system as is the case in HalfLife 2′s “Source” engine.

I like the cute 3D hands on the floor that presumably denote all the finger controls, but still prefer to have those controls seperate in a Channel Box (Maya) or Modifier Panel (Max). Note the boneless cloth garment, showing that they have a seperate system for that element rather than being bone-driven.

What really intrigues me though are the long bones at the extremities – possibly something to do with their real-time IK system, but more probably with the standout feature – the deforming collision of the Colossi themselves, whereby they calculate the player’s position on the collision as a single spherical point making it easier to react to the deforming mesh, (below).

The aforementioned real-time IK system appears to support a fairly standard 2-bone setup (affecting just the last two in the horse’s leg), save that it not only affects the limbs but also raises and angles the whole character in one solution. In past games I’ve encountered these as two seperate systems, with the latter affected by a check on the ground normal rather than an average of the various IK limbs. See below for a comparison – first without, then with the root being affected.

Finally, the most impressive (and hopefully soon to be widely adopted) element of the entire movement system centres around what they call their “motion-addition” system – essentially the combining of procedural and animator-created motion to give true fluidity without the sloppiness of purely-procedural movement, or rigidity of keyframes alone.

A physical simulation drives animation of the bones, causing the character to rotate around his single handhold, and presumably to arc the body and delay the limbs based on that motion as seen in the image below. This was not only used to afford the flailing motion of the player character while the Colossus violently attempts to shake him off, but also for the natural movement of the horse.

This combination of procedural calculation that drives motion tweaked by an animator, (or even better, purely procedural systems that are simply fed data in the form of exported animations) is the next step in more natural-feeling motion in our in-game characters.

“Movement becomes wooden if purely controlled by a program, but also a pre-made animation won’t fit in right. It is the combination which makes it work. When I saw this working in real time, I was more than a little impressed! I think it has become a truly wonderful system to combine the animator and programmer.” - Fumito Ueda, Director

For the original article in its entirety (including images), go here, with an English translation here.

Building on the stealth-based reliance on gadgetry and darkness of previous installments, Chaos Theory offers a more open-ended approach to each situation, populating the less-linear levels with more forgiving (or fallible) AI, losing their super-human eyesight, allowing the player many more creative ways to dispatch of each enemy (or not) as they see fit.

Gameplay aside, Chaos Theory features the most advanced realtime IK system the alsoran has witnessed in a videogame so far, making player character Sam Fisher convincingly connect with the environment and pushing the medium ever closer to true procedural movement that is entirely under the command of the player.

IK On Inclines

IK is used extensively in this game, allowing player character Sam Fisher’s feet to convincingly conform to inclined collision on stairs and ramps (though ramp collision is used for stairs also) whilst walking and standing still.

In order to avoid hyper-extension of the IK limbs, all driving and idle animations have been created with Fisher’s legs slightly bent, though this can also be achieved by procedurally dropping the root-bone every frame, depending on the lower of the two IK feet, or the incline angle taken via a normal-check.

Direction Changing

IK is used not only in the up/down axis, but also to prevent the feet slipping and rotating as Sam changes his direction. This is prehaps the simglemost difficult area to prevent unnatural movements due to the player’s ability to change direction at any point in a walk/run-cycle, regardless of which foot the PC is on, which would greatly affect a real person’s balance.

As it would be undesirable to wait for the correct foot to come into contact with the ground before allowing a weight-shift to the new chosen direction, Splinter Cell’s developers have come up with an aesthetic solution to lessen the effect of the still-present unnatural weight-shift by locking the feet’s position and orientation during the shift to the new direction.

This creates a subtle, almost unnoticeable, re-affirmation that the player character’s feet are firmly taking the weight during the course of any unpredictable movements the player may perform, but it is with this subtlety that game animation slowly creeps towards more believeable motion that does not jar the player out of the experience.

Momentum-Shift Weight-Taking

The crowning glory of Chaos Theory’s IK system, however, goes to the advancements in Fisher’s weight-taking as he moves from idle to walk and vice-versa.

Typically when the player releases forward movement on the joypad thumbstick, the player character will either procedurally transition into the idle animation, or play a walk/run_stop animation created to show the shift of weight and foot movement required during this (often immediate) change in momentum.

However, chaos theory appears to use an advanced combination of foot and root-bone IK to hold Fisher’s current pose – perfectly apt should he be required to “freeze” while sneaking up on his quarry.

The simplest way of achieving this is to rapidly slow (or immediately stop, depending on the PC’s momentum) the playback of the current run/walk/sneak cycle so it holds the current frame of the animation.

Next, the IK of any foot mid-stride must be dropped to the floor, maintaining its horizontal position. Finally, the root-bone position also drops to provide the appearance of Sam taking his weight, (in non-duress situations only), and assuming a more comfortable position rather than eternally holding an unnatural mid-stride pose.

The Ubisoft developers have even had the foresight to provide a little overshoot on the procedural root-drop, giving the action a natural bounce that would otherwise appear mechanical were a simple linear transition to be used.

Importantly, the sneak action’s playback must resume from that same held pose again to provide a smooth transition out of the idle.

Player-Created Posing

All these actions combine to produce a completely variable, and importantly, player-defined idle pose. Be it whether Fisher is standing on uneven terrain, or at which point he has broken from his stride (or both).

The developers even show off with the added ability to always connect Fisher’s left hand to his leg as he settles on his haunches – very impressive indeed.

N.B. It should be pointed out that several of these, and other techniques were pioneered (or at least first witnessed) in the technically astounding but criminally undersold Transformers on PS2 from Australian developers Melbourne House, though the alsoran could write multiple articles on the technical merits of that game alone.

Conclusion

These are just a couple more examples of moving towards more convincing character movement in videogames. By eliminating unnatural occurences such as floating feet and repetetive idles, that “suspension of disbelief” is that little bit easier to maintain.

This does, however, reinforce the alsoran’s belief that in a move towards realistic characters lies a move away from twitch gaming and instant PC response, as that’s a context under which the Splinter Cell series has always thrived.