QUOTE(L. Spiro @ 01/23/09, 12:43 PM) [snapback]299476[/snapback]
Or if I sample every frame as a keyframe and do linear interpolation, is it basically smooth enough that there is no difference at all?
Yes atleast as smooth as the maya operator sees. Be sure t interpolate the rotations by not using eulers. This is teh best you can do in any sane timeframe without neglecting all the ings you cant approach with this strategy
But how the anim curves work is described in teh api manual, their EXACT interpolation is there (so you can implement that but it takes you nowhere, and no theya re never linear its just a special case of the same spline). But thing is theres a lot more things than just IK that desnt get out this way. For example you would be loosing out n 99% of everything we use in maya. No expressions no sdk's no connections no path anim.
All of theese are very common in maya rigs in fact if if compare mys tandard rig theres lesss controlled by keyframes than any other control. And even ik is just the tip of the iceberg.
The only sane way is to step toug time with some interwall and fit teh data somehow or pray yur interpolation doesnt alias.
QUOTE
But in both cases I would still need to know whether to use linear/spheric quadratic for rotations and linear/hermite for vectors.
Maya never actually uses any of those. It just uses the same spline everytwhere except in rotations when it interprets the data as quats or sychronized eulers or independent channel euler values.