time shift by point/prim attributes?

[willlinks]

Hi,

 

I'm tying to use point or prim attribute to specify frames to offset the time for each primitives animation.

Is there way to feed the attribute in TimeShift?

Or is there work way around without using TimeShift?

 

I need to offset attributes not only P but also other attributes.

 

Thanks in advance for your help!

 

 

[anim]

not with TimeShift

use CHOPs for this

[willlinks]

Tomas,

Thanks for your reply!

 

The only problem with CHOPs is that it will take a while to load all frames...

I can also use the copy stamp to feed prim attributes in SOP but it also takes time...

[anim]

what you want to do requires accessing attribute value at different time than current

that requires cooking geometry at that time

if you have 1000 points/prims and each of them forces geometry to cook at different time it would be 1000 individual cooks to get your data even if that time range is only 1 frame long (as it may be 1000 individual subframes)

and that happens every frame so for 100 frames it would be 100000 cooks of your heavy geo

so CHOPs approach looks the most optimized solution

if you let CHOPs to cache and let's say your range is 100 frames, then you wait just 100 cooks of the geometry, you can save it to disk or just lock the Geometry CHOP

then further processing has all the data needed for that range and you can sample any time within for any point necessary pretty fast

so no need of cooking source geo anymore

if you can't fit all range in the memory at once in CHOPs, you can analyze what range of time is necessary to sample for current frame and load just that range in CHOPs, that will need to recook geo at each frame of that range for each processed frame of the animation, so you loose a bit of speed, but would be still faster than bruteforce cooking geometry for each point

the more optimised solution would be to cook and cache only integer frames that encompass the requires time sample for each point/prim and interpolate between, which would be the same as CHOPs method, but more efficient for cases where time samples are not present alongside the whole range per frame

[willlinks]

Hey Tomas,

 

You are right.

I'll go for CHOPs approach then.

 

Thank you so much for the explanation!