[sldev] Body motion and facial expression tracking, Microsoft did it

[Dale Mahalko]

In general I expect that scripted interaction animations have to be
somewhat unreal to deal with issues of latency and alignment. Neither
participating avatar can be expected to be aligned "just right" to
perform any given action on the fly

So they both virtually step forward to align with the scripted motion,
perform it, and then step back to their original position... even
though on the sim side neither avatar actually moves.


Due to the issue of limb positions sinking inside for prim-bodied
avatars, the existing rigid-position animating isn't good enough,
since a given handshake animation may look all wrong applied to a
prim-bodied AV vs a standard AV.

If you want to get serious about it, the motion system needs to have
adjustable min/max positions for where limbs can be located relative
to the body shape, practically doing physical collisions of say
elbow/wrist vs prim-body surface and knowing limits of arm
rotation/bend, and having the whole avatar move closer or away in
situations where the movements are either out of range of normal arm
motion or are causing arm-to-body collision.

How does an adult shake hands with a child? The arms of the child are
too short so the adult is required to either bend over at the waist or
bend the knees, or both. Can your interaction solver handle this
extreme situation?


Though it's a total joke to discuss this in relation to the current
avatar animation system. Arbitrarily tall or short avatars are not
permitted. The avatar needs to be weirdly distorted to make tiny AVs,
so there IS no hand/arm/elbow to do interaction with, on a tiny
avatar..

The existing avatar animation systerm would have to be thrown away and
replaced with something that really does allow avatars down to 10 cm
tall, with 2 cm long appendages, to allow dynamic direct-calculated
AV-to-AV interactions.

I'm not volunteering for this. :-)

- Dale Mahalko / Scalar Tardis