Biomechanical¶
The engine for characters people look at up close. How it keeps its gaze intent apart from the body carrying it out, how the eyes and head share a look, and what it does with a target it cannot reach or cannot see.
Gaze Intent and the Body Are Kept Apart¶
A classic look-at computes the pose that would point at the target and moves the bones toward it. Biomechanical does something different. It keeps what the character is trying to look at and what its body has managed so far as two separate things, and it drives the body by the difference between them, continuously, the way the real system does.
Because the intent and the body are separate, looking away does not erase the intent. The character still knows where the target is, so looking back is a movement with a beginning and an end, not a snap to a remembered value. A target that moves is followed, and a target that stops is held. And the head keeps turning until the eyes have recentred in their sockets, which is the thing that makes a head turn read as a person turning to look rather than a rig rotating.
Built From the Research¶
The engine is put together from the published science of how people look, and its inspector is laid out the way a look happens, as three responses in the order they occur.
- Cognitive Response is what the character works out: where the target is, whether it is moving, and whether the look is done yet. It follows a moving target the way a person does, by predicting where it is going, so it keeps up through a moment the target cannot be seen.
- Physical Response is how the body moves. The head has weight and swings like it, the eyes settle the way real eyes settle, and a look of any size takes exactly the time a real one would.
- Reflex Response is what keeps everything steady while all of that moves: the eyes stay on the target while the head is still turning, and the head steadies itself in space and comes back to the body line.
Beside the three, Physiology holds the character's built-in traits, where the eyes settle when they have nothing to do and how far the peripheral vision reaches. Those are the character. The responses are its behaviour.
How the Eyes and Head Share a Look¶
When a person glances at something nearby, only the eyes move. When they look at something further round, the head joins in, and for a big turn the body follows. Nobody decides that consciously, and nobody tells this engine either.
It works the same way the real system does. The eyes take the look first. The head joins only once the eyes would have to reach further than is comfortable. Because of that handover, a glance and a full turn come from the same rule, and every size between them lands where a person's would.
Where the Head Stops¶
People leave a little of a look to the eyes rather than turning the head all the way, and that slight lead is a large part of how a look reads as alive. Stop Short Of The Target, in the engine's The Gaze section, is that behaviour. On, the head stops short and the eyes carry the rest. Off, the head turns the whole way and the eyes centre, for a character that should square up on whatever it looks at.
Takeover Endpoint says how the stopping point is authored. Eye Residual keeps a set number of degrees for the eyes at the end of every look, small or large, which is how real eyes behave. Fraction gives the head a share of the whole look instead, which is simpler but leaves the eyes far off centre on a big turn. Head Takeover (%) is that share, read only under Fraction.
Settling Over Time¶
Settle To The Target Over Time, in the same section, lets a held look gradually complete, the head coming the rest of the way round and the eyes returning to centre. It is a stylistic settle rather than something people reliably do. It waits out a hold first, takes its time, and never carries the head past its comfortable range.
Comfort & Reach¶
A target beyond where the character can comfortably reach is handled, never ignored. The character reaches as far as it comfortably can and holds, and the strain of holding is itself relieved over time, the way a person at the edge of their vision eases their body round. Beyond that, the character can report to your game that a target is out of reach and by how much, so the body can be turned by the thing that owns the body.
A target the character loses, one it can no longer see and that stops being refreshed, is coasted on for a while and then given up, rather than stared at forever.
Comfort Limit¶
Comfort Limit, in The Gaze section, is what the head's comfortable range means for this character. It has three answers, and the first is the one described above.
| Comfort Limit | What the head does past comfort |
|---|---|
| Retreat | How people behave. The head holds the look for a while, tires, and eases back. |
| No Comfort Limit | The head goes wherever the look asks, out to its mechanical limits, and never pulls back. A character that never tires. |
| Limit Inside Comfort | The head never leaves its comfortable range at all. It reaches less far, so more of a deep look is left to the eyes. |
Beyond Reach¶
What the character does about a target it cannot reach at all is in the Worn Hold, Beyond Reach section. Out Of Reach is the conduct: Regular settles at the edge of reach, holds with endurance, and relieves when the hold wears out, while Eyes Only leaves the head at rest and lets the eyes carry their own share, a character that will not crane.
FaceCue never turns the body itself. Where a character stands and which way it faces belong to your game, and this is where FaceCue hands over. With Request A Body Turn on, the moment a target is found to be beyond the reach of eyes, head, neck and trunk together, the Eyes driver raises the BodyReorientRequested event, and raises it again if the character lost track of that target and then catches sight of it again. It carries a BodyReorientCue with how far the body would have to turn, sideways and vertically, where the target is as a bearing from the character's current facing, and the whole angle from body-forward to the target. Hook it to your character controller or animation system, and let that decide whether the character turns, steps or stays where it is. Turn by the reported deficit and the target lands back inside reach, and the request closes on its own. With the toggle off, or nothing subscribed, the character settles at its reach silently. The event's fields are tabled on Look Target.
A Target Behind a Wall
Your look-target source carries a flag saying whether the target can currently be seen, and it is yours to set: a line-of-sight check, a trigger volume, whatever your game already knows. Say a character is watching someone run across a street, and the runner passes behind a parked van.
While the runner is out of sight, the character keeps looking where the runner would be by now, along the line it had been following, exactly the way a person's eyes keep travelling along the path they expect. If the runner zigzagged behind the van, the character is briefly wrong, and the moment the runner comes back into view the look corrects with a real movement to where they are. A typical look-at has only two options, and both give the game away. It can stop dead the instant the flag drops, so the character freezes on the van. Or it can go on tracking the real target through the van, so the character follows someone it cannot possibly see. Neither is what a person does.
When to Use It¶
Biomechanical is for characters people will look at up close. That is the short answer, and the rest of this section is what it means in practice: where the difference is visible, what the engine needs from you, and when to leave it alone.
Where the Difference Shows¶
The engine earns its cost in the moments a person would notice a rig.
- Eye contact at close range. In VR the player stands where a person would, and looks a character in the eye. This is where the eye-and-head split is plainest: a glance at the player is eyes only, a turn to follow them round is the head, and the handover between the two is what makes the character read as present rather than tracking. Set Personal Space on the Eyes driver and a player who steps in too close gets a lean back and a release, the way a person makes room. See Too Close.
- A held look. Dialogue, a listener, anyone whose look rests on one target for more than a moment. A look-at that only aims reaches the pose and stops. This engine settles, and a look held at the edge of comfort is eased over time, so a long conversation does not look like a freeze frame.
- A moving target. A player walking past, a character crossing the room. The look is led by where the target is going, not dragged behind where it was, and a target that passes out of sight is followed along its path and picked up again when it reappears.
- A body that joins in. A seated character behind a counter, a character who should turn in place to follow the player. The posture options say how much of the body is free to turn, and a target beyond reach is reported to your game so the body can be turned by the thing that owns it. See The Body.
What It Needs From You¶
Less than it sounds. The engine's model is inside it, and the defaults are the measured ones, so a character set up by the OneClick Wizard is already running it. What it takes from you is the same as Simple takes:
- A head bone, found at setup, and a chain below it if the body should join a look. Leave the chain empty and only the head turns.
- A target, set through the Look Target input, and the visibility flag on it if your game knows when the target cannot be seen.
Everything else is tuning, and most characters never need it. The three responses and the physiology are there to change a character's temperament, not to make the engine work.
When Not To¶
For a crowd, or a stretch of heavy action where nobody is watching faces, the engine is doing work nobody sees. Simple gives the familiar game-character look at a lower cost, and both engines share the eyes at rest, the blinks, the pupils and the idle motion, so a crowd on Simple and a lead on Biomechanical still look like the same kind of character. How much Simple saves, and why the output rate is usually the better lever, is on the introduction.