hey guys!

I have a quick question. So, I am going through some houdini modeling tutorials, and I am watching them, and noticing that whenever someone in the tutorials is modeling an asset there seems to be no way to commit all the nodes laid down for that asset.

to elaborate, lets say I am modeling something in maya or xsi. Well as I add operators/operations to my geometry the history grows. In houdini the network that gets built grows. But once one is finished with the model and they know no further changes are going to be made, is there a way to commit the model to a final form free of any history?

or

is this just not the case with houdini? Is this some of the powerful traits that houdini has to make it non-destructive and truly non-linear?

thanx

john syracopoulos
fx artist
montreal

The concept of committing your history is, alas, baggage from other packages.

The beauty of *true* proceduralism(as opposed to the partial implementation you get in other programs that name-drop that term) is that you don't need to really think like that. You can do the equivalent of it anytime you want, at any point in the chain by simply throwing down a ROP Output Driver SOP and writing out the geo. Insert a File SOP to read it back in again, and you've got your save history. Obviously, you could read that back in an asset too.

Additionally, you can lock any SOP, which saves the actual computed data at that point in the chain in your hip file. Be careful with this, it can bloat file size.

In my books, there is no final until the week after the job has delivered.

You can pick and choose where you want to commit or even if you really need to, based on whether or not you need proceduralism, and speed of the network.

Cheers,

J.C.

This is definitely doable in Houdini. You can either toggle the Lock flag and delete all the nodes above your end one in the chain, or collapse them all into a subnet so they are still around for later use. You can also write the geometry out to a disk file then read it back in and delete the chain. This can be preferable if your geometry is heavy because if you lock the Sop that geometry is embedded into the hip file and makes its file size large.

Edit: JC beat my slow iPhone typing skills

thanx guys

If you just lock your SOP, there's no reason to delete anything.

you can also right click a node and save geometry if you're only doing a single frame.

JColdrick
The concept of committing your history is, alas, baggage from other packages.

The beauty of *true* proceduralism(as opposed to the partial implementation you get in other programs that name-drop that term) is that you don't need to really think like that. You can do the equivalent of it anytime you want, at any point in the chain by simply throwing down a ROP Output Driver SOP and writing out the geo. Insert a File SOP to read it back in again, and you've got your save history. Obviously, you could read that back in an asset too.

Additionally, you can lock any SOP, which saves the actual computed data at that point in the chain in your hip file. Be careful with this, it can bloat file size.

In my books, there is no final until the week after the job has delivered.

You can pick and choose where you want to commit or even if you really need to, based on whether or not you need proceduralism, and speed of the network.

Cheers,

J.C.

I dont quite understand this. If I throw down a ROP then I get render parameters. I see an option for outputing the geometry but couldnt quite get it all to work. Is there a tutorial or any existing online literature on how to perform this/similar actions?

johni

So there's no confusion, that's the SOP I'm talking about. That will output the data fed into it at $HIP/$F.bgeo(startup directory, with the frame number as the name). That is geometry as cooked by the network up to that point. You get the advantage of the ‘Save History’ mentality without having to delete anything - you can always go back and tweak your network, and resave out the geo.

Again, unless you're getting into heavy complicated networks, you typically won't even bother doing this. It is invaluable in production, though, for the compute intensive stuff.

Cheers,

J.C.

- I'm not 100% that Apprentice has this SOP - might be a limitation. Unsure. Anyway, it's in the production version, and as others have mentioned, you can always right click/Save Geometry…