Cookie trouble??

   5996   11   2
User Avatar
Member
289 posts
Joined: July 2005
Offline
Hello everybody,

I'm trying to make an animated boolean(union) with turning an moving boxes.
6 boxes in total. But I cannot get it stable for the animation. several frames have bad booleans, parts missing, or completely missing.

After each boolean I make sure it is one closed object again using the fuse. but it does not help a lot.

Any idea's? Because it seems really very unstable.
A computer is almost human - except that it does not blame its mistakes on another computer.
User Avatar
Member
12469 posts
Joined: July 2005
Offline
I find that you must turn on jitter in the Cookie SOP and also Divide your cubes into triangles. This will give you better results, generally.

Hope this helps,
Jason
Jason Iversen, Technology Supervisor & FX Pipeline/R+D Lead @ Weta FX
also, http://www.odforce.net [www.odforce.net]
User Avatar
Member
289 posts
Joined: July 2005
Offline
thanks Jason,

but it does not help a lot, it is a bit different but not a real solution to my problem. Have you any other ideas?

Cheers
A computer is almost human - except that it does not blame its mistakes on another computer.
User Avatar
Member
710 posts
Joined: July 2005
Offline
That's weird, jitter should work. I'm also using the Cookie SOP right now in an animation, and using the jitter I'm still getting 8 bad frames out of 175, but simply changing the seed on the jitter fixes those. Are your objects dense enough? I think you get better results when your geometry isn't too low-res.
User Avatar
Member
2199 posts
Joined: July 2005
Online
I did see a tip once that suggested to get the best results from a cookie always put the simplest piece of geometry into the left input. Maybe worth brickering the cubes too first using the divide sop, and double check all your primitive normals are pointing outwards.
The trick is finding just the right hammer for every screw
User Avatar
Member
289 posts
Joined: July 2005
Offline
Thanks for the replies!

Hi Dajuice,

It showed some result but no real improvement, the problem is that what I want to test with it does not allow manual tweaking. It should be able to construct loads of “random” generated objects created out of booleans with some boxes and spheres.

Hi Simon,

The divide together with switching the simplest geometry into the left input gives the best result until now. But still not stable. +/- 30 frames are wrong on 250.

Are there improvements on the cookie sop in newer versions? I'm using the apprentice version 7.0.231.

Else I think this is still an important issue to have a look at for sidefx.
A computer is almost human - except that it does not blame its mistakes on another computer.
User Avatar
Member
2199 posts
Joined: July 2005
Online
It's odd that you are having so many problems. The Cookie sop was the “stable” replacement for the very “unstable” Boolean sop. I use it tons with no problems what so ever. Although generally for modelling, so perhaps the animation is throwing it out.

The only other suggestion/question is what scale is your geometry, ie are your boxes unit size or 1000's of units? This will effect how the jitter performs, best if they are closer to a unit or the jitter amount has to be much bigger.
The big thing to watch out for is coplanar faces it really doesn't like them, so if you have lots of boxes the same size and they are all oriented the same way this will cause problems, maybe try randomly scaling them all before the cookie.
Lastly try mergeing all the transformed ones together and then just doing one big cookie with the original in the left input and the merged ones in the right.
The trick is finding just the right hammer for every screw
User Avatar
Member
289 posts
Joined: July 2005
Offline
No the scene is nothing special I think.

I have uploaded the scene so perhaps if you could have a look at it of what I'm doing wrong.

There are 2 outputs I want to use here. clip1 and clip2 both of them are unstable.

http://users.pandora.be/mantis/CookieProb.hipnc [users.pandora.be]
A computer is almost human - except that it does not blame its mistakes on another computer.
User Avatar
Member
2199 posts
Joined: July 2005
Online
Hmmm…. I see your problem, that's a mighty complex boolean.
I can't get it to be 100% stable either. Actually I amazed it works as well as it does. Why do you actually need to boolean them all together?

The only thing that perhaps helps a little is to put all the divides before the transforms so that the boxes aren't getting redivided every frame, but the real problem is that the geometry is so random and changing so much you are almost bound to hit a few frames where the geometry can't be cleanly resolved. Perhaps Sesi can improve the code but I tried it in H8 and it has the same problem, I think it is asking a lot to get this to work. Sorry.
You will probably have to hand tweak the solution for the frames that don't work. Can't see any other way around it.
The trick is finding just the right hammer for every screw
User Avatar
Member
289 posts
Joined: July 2005
Offline
Okay thanks Simon!

Npowertools has created a very stable and very powerfull plugin for doing just this.
perhaps sidefx can pursuade them to make the plug available for houdini as well.

I use it quite often and it is really very stable!

http://www.npowersoftware.com/booleans/learnmorePB.html [npowersoftware.com]

examples







A computer is almost human - except that it does not blame its mistakes on another computer.
User Avatar
Member
2199 posts
Joined: July 2005
Online
Yeah I've seen this before, it looks like it makes very clean geometry, far fewer triangles. However I still like to see it do a scene like you have created. I think any one of those example renders could be done with the cookie sop as they are just static.
Since you have it already why not try doing your Houdini scene with it?
The trick is finding just the right hammer for every screw
User Avatar
Member
289 posts
Joined: July 2005
Offline
hmm, your right. It is a bit more stable but not perfect. I had to leave out some geometry to make it work as well.
A computer is almost human - except that it does not blame its mistakes on another computer.
  • Quick Links