Thanks everyone for your insight and suggestions! Raffaele: In this particular case I wasn't so bothered with the aesthetics of the blur, but the wheels seemed to flicker in and out of existence, some frames the car would appear to hover, others the wheel looked extremely stretched in one axis, sometimes I get the arc, and in other shots it would work just fine.
Enveloping the wheel always produce the correct result, even with silly speeds, so the whole thing would go away if MR could just be forced to evaluate motion per-point instead of using object transformations, as an option. It's not even slow actually, so this particular optimization just creates more problems than in solves. Is this something that could be controlled by the host application, or is it strictly a Mental Ray issue? Seems like Soft should be able to send vectors per-point to MR? On Thu, Jul 19, 2012 at 2:29 AM, Raffaele Fragapane < [email protected]> wrote: > It's not an uncommon problem in rendering in general. > When we do things like wheels and propellers we normally render wedges of > various settings at different incident angles, speeds and so on to make > sure the blur looks correct. > > In these cases the physical accuracy of the motion is absolutely > irrelevant (IE: the wheel is theoretically slipping on the ground all the > time or the propellers is at a fraction of the RPMs required for the plane > to move at that speed), the quality of the blur is everything to sell those > things off. > > The B52 in Sucker Punch took a full two days of testing and rendering > wedges and an operator that would modulate the RPMs based on aesthetic > choices to produce usable frames. > > Another classic issue is emulating the forward + rewind + stable blur seen > in footage of car wheels accelerating past a camera, which we had to > implement controls for so that people could animate parameters based on the > actual aesthetics of the blur and motion they wanted instead of animating > rotations and shooting in the dark. > > This was the case with PRMan and Mantra and was run with all kind of data, > from straight point caches with a stupid amount of subframes to straight > transforms to deformation with additional data coming from a monitored > transform. > > Not that it's not an issue or that it shouldn't be looked at, just saying > that if you want to do spinning objects and you want that cinematic blur > people are used to you'll have to bite the bullet and NOT go for temporally > or physically accurate. > > Even shooting these things for real on a set often requires tests and > wedges for the rigs to be timed and controlled so the DoP gets what he > wants. > > > On Thu, Jul 19, 2012 at 9:44 AM, Jack Kao <[email protected]> wrote: > >> I wonder if it's something to do with Mental Ray specific? >> > > > > -- > Our users will know fear and cower before our software! Ship it! Ship it > and let them flee like the dogs they are! > >
