If moving between resolutions is a pain then you can always have visible low-res geo with the high res stuff hidden. You don't get the same speed boost that you get with multiple resolutions, but sometimes it's enough to make it workable.
On 23 December 2013 18:33, Cristobal Infante <[email protected]> wrote: > another way is to use 2 resolutions for your reference models. One very > low res, and only switch to the highres when you are ready to render. The > trick is to use the same rig so the animation sticks. > > If you want to give this a shot I recommend you download the scn toc > manager created by Alok to switch those models from outside the scene. > > The standin route is great if you are rendering with Arnold IMHO. > > good luck! > > > > On Monday, 23 December 2013, Paul Griswold wrote: > >> Thanks Alan! Yep, this is all solid, non-deforming stuff (CAD data). I >> have rigged it all with nulls the way you mention. The problem I'm having >> is just updating when I scrub the timeline. Even if the object has zero >> animation, simply scrubbing the timeline causes Softimage to freeze for >> several seconds. >> >> I'm actually testing Redshift for this one. They've got a stand-in, but >> it's very alpha IMHO. You either get a box or a full representation of the >> geometry. There's nothing in-between, yet. VRay's implementation looks >> somewhat nifty - I think they give you a bunch of points in the shape of >> your object. >> >> It sounds like the fastest solution would be to decimate the geometry and >> use it as a stand-in. >> >> I just wasn't sure why there's such a huge difference in performance >> between local and reference. It's really significant. >> >> Thanks! >> >> Paul >> >> ᐧ >> >> >> On Mon, Dec 23, 2013 at 1:05 PM, Alan Fregtman >> <[email protected]>wrote: >> >>> >> Is it as simple as building low res stand-ins and offloading? >>> >>> Pretty much... yeah. >>> >>> Is it a solid, (mostly if not completely) non-deforming thing like a >>> vehicle? If that's the case, you may wanna consider making a null hierarchy >>> where each null contains a selection of meshes that "move as one". >>> >>> For example, a regular solid car's rig might be: car body, left/right >>> front/back wheel rotation and left/right front/back wheel brakes, so 9 >>> nulls. Animating those 9 nulls will be way lighter than dealing with >>> hundreds or thousands of parts deforming or individually constrained to >>> whatever, plus it's less data for the Delta property to keep track of. By >>> the way, I like to call these nulls "segment nulls". >>> >>> If you're dealing with mentalray or Arnold, both have the standins >>> concept that works quite well, especially in Arnold. (Maybe XSI Vray does >>> it too, not sure.) >>> >>> You'll want a standin per "segment" and if you name your standins the >>> same as your segment nulls in a separate resolution, then it's very easy to >>> animate a very light rig that is high-res compatible. Also makes it a piece >>> of cake to republish update geo and shading by simply reexporting the >>> standin files. >>> >>> At work we had stupid mesh density in Pacific Rim's control pod "stilts" >>> and this *segmented workflow *of standin nulls constrained to a rig >>> worked out great. ;) >>> >>> >>> On Mon, Dec 23, 2013 at 11:09 AM, Paul Griswold < >>> [email protected]> wrote: >>> >>>> I don't work a lot with reference models, but I am now. The mesh is >>>> dense and has a lot of parts to deal with. When the mesh is local, >>>> Softimage handles it fine. But when it's reference, just moving through >>>> the timeline takes 8-10 seconds per frame. Even in Bounding Box mode, >>>> Softimage grinds to a halt. >>>> >>>> Can anyone point me to any FAQs or guidelines on working with heavy >>>> reference models? >>>> >>>> Is it as simple as building low res stand-ins and offloading? >>>> >>>> Thanks & Merry Christmas! >>>> >>>> -Paul >>>> >>>> ᐧ >>>> >>> >>> >>
