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

Tateru Nino tateru.nino at gmail.com
Sun Jun 7 08:54:39 PDT 2009



Argent Stonecutter wrote:
> On 2009-06-07, at 08:07, Jan Ciger wrote:
>    
>> This is also why the simplistic approach with "only modifying existing
>> animation and locking/weighting down joints" as Argent described
>> wouldn't work - if you lock down the feet ("I am only reaching for
>> things!"), you either wouldn't be able to find the solution (with the
>> VRlab's solver and the balance constraint on) or it would look
>> completely wrong once the object is out of arm's range and you need to
>> bend over to get it.
>>      
>
> That's WAY more sophisticated and computationally expensive than I'm
> suggesting.
>
> Imagine a canned animation having the avatar drinking a beer. The
> avatar is siting on a barstool, and he's wearing the beer and a target
> prim in his mouth. His upper body bends back slightly and his arm
> lifts. When his hand (holding the beer) gets within a foot of the
> target prim the preloaded override provided by the script starts
> adjusting the elbow and shoulder by a few degrees... making the motion
> move the beer to where his mouth actually is, instead of 5cm inside
> his head.
>
> That's the problem I'm interested in solving. not the much harder
> problem you're talking about.
>
> In your example, if the object is out of range nothing would happen.
> The existing animation would run... so the worst case is no worse than
> what we have now.

So, if I've got this straight, the problem is essentially two-fold:
1) Something (probably the viewer) should ideally be solving for 
specific animation outcomes (possibly by adapting template/guideline 
animations) and,
2) There's currently no actual way to specify what those outcomes are.

-- 
Tateru Nino
http://www.massively.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.secondlife.com/pipermail/sldev/attachments/20090608/91f0d2f0/attachment.htm 


More information about the SLDev mailing list