And i would never edit data in the spreadsheet…

Hehe - yep, feels wrong, doesn't it But it really is the best way to store data like this:



The nice support folk have suggested that I store this info in an external text file, then use something like the Table Load SOP, which sucks in a CSV file, creating points and attributes all at the same time. Which is grand, but doesn't lend itself to quick iterating and tweaking, which really is what I need when developing both my HDA and my “running order”.

I've raised an RFE asking for a Table Load SOP that just uses Houdini's own spreadsheet, rather than relying on an external package. If they do add it as a new feature, it'll certainly help my workflow; it'll let me populate my scene with a few instances of an HDA with varying parameters so I can tweak the HDA and see how it looks in different situations.

Thanks for the tips, though Olaf - much appreciated

Yes - but that doesn't really help here, as most of the instances of my HDA will have different parameters set. Plus the meshes output are small and lightweight so performance isn't an issue in this project.

S'pose I'm asking: if you'd made a firework HDA, configurable to generate every kind of firework your show needed, where / how in Houdini would you actually design your show?

It's really a question of how best to store the data for all the instances. Point attributes seem the most sensible, and they're easiest set in the Spreadsheet, but … using the spreadsheet always seems a little fragile somehow.

In an ideal world, we'd have a specialised point generation SOP - a “Create Points plus Attributes” SOP that'd let you create and edit points (rows) and attributes (columns) in the spreadsheet without destroying what was already there.

Edit: I've logged it as an RFE.

I'm doing a section of a music video: I've made a simple “blooming petals” HDA with a number of parameters, and now I need to set up lots of instances of it that fire off at certain beats and cues within the music.

If I wanted to orchestrate a fireworks display, I'd have exactly the same question - I could create a super flexible über-firework HDA, but I'd then need a way to store the parameters and timing for each instance.

There are a number of ways I could do this, but I'm interested to know what others feel is the best approach.

My HDA has a pretty simple interface:


And I need to set up / orchestrate around 50 instances. So I could just make lots of copies, tweak their parameters for timing and taste, and merge 'em all:


(yuck)

But it'd be much nicer to be able to use the spreadsheet to set all those parameters and delays, so an approach like this, using points and attributes may be better:


At least I can see and tweak all the parameters in the spreadsheet nicely. That locked “All_the_orchestration_information” Edit SOP lets me see all my lovely data:


But it has its issues. If I want to add another parameter to my setup, I have to unlock that Edit node and lose all my orchestration. Or if I didn't create enough points to start with… things could get messy.

What are your thoughts? Is there a better way to achieve a spreadsheety display of my timings and parameters? Is it necessarily evil to edit in the spreadsheet like this?

If I just want to grab a material or a single object from another scene, seems the only way to do it is to save the current file, load up the other one, delete everything but the one object/mat you want, save that under a new name, reload your current file again, then merge.

Is there a quicker way?

Gah - this is happening to me more frequently now, too. Usually seems to be caused by me clicking a parameter value, typing away, then noticing my mouse has wandered into a different pane. Result: one of my network editor panes seems to have keyboard focus locked to it.

Might try and break things on purpose later to try and track it down.

I'm a little confused. Is the problem that you're scattering before you fracture, so after the fracture all your sourceprim / sourceprimuv attributes no longer refer to the right prims? Any reason you can't scatter after the fracture?

Or perhaps you could use another attribute (@Cd perhaps) to store group information during the fracture and then restore it later?

I'm not entirely clear on what you're trying to do. Can you upload a (simplified if necessary) project file?

OK, I have dangerously little knowledge, so this could be horribly wrong, but hey

You only need the pinned points to be updated each frame, so I split the constraint networks in two - back out in SOPS land, instead of merging the pinned/sprung constraints and then piping them into the DOPs net, I wired them into two separate inputs. Then inside the DOPs net:


Now the springs don't get overwritten every frame (but then they don't need to be, once things get going, ie after frame 20ish). The pin constraints still follow the sphere… all seems to still work.

And there my expertise ends.

but I've learnt a ton looking at your project file, so at least one of us got something out of this eh

Man, looking at those images, you've got a lot of points there, huh Looks like a point cloud from a LIDAR scan.

The techniques ndickson was talking about are grand if you're repeated structures going on, but I'm guessing you haven't built your scene up from smaller components.

There are soooo many points in your scene though that any sort of instancing is going to eat memory. But how many are actually going to be close enough for you to see they're boxes and not spheres? I can only guess at the shot you're after, but I'd be tempted to group only those points within a certain distance of the camera, and only instance boxes onto them, leaving the rest to be rendered as spheres or discs.

Level of detail stuff, innit.

Quick example attached of grouping points by distance from camera. I'm grouping by Bounding Sphere, and to tie the centre of the sphere to the camera, right-click the “Center” parameter, and choose Reference -> Scene Data, then drill down into the camera til you reach it's World translation. Houdini fills in the zany expression you need automatically

Move the camera around and the point colours will change automatically. You can object merge either the near or far ones into a new object (or instance object) as appropriate

This crashes my Macs, but they're not of the best pedigree - could someone check if this crashes their machine too, before I report it? It's cost me a couple of hours tracking it down this morning. Made a simple test scene though:

Two boxes, one with points scattered on it: feed them into a Tetrahedralize SOP. All fine, unless the boxes happen to match dimensions in any one axis, which makes Houdini crash out on all 3 of my machines.

Ta!

Hehe - well, my old approach was simplistic but it wasn't quite as crude as copying files to all the machines

With project files / textures / render bins all on network drives, setting the whole “farm” going was as simple as VNCing into each machine, double-clicking on the project file and hitting “Render”.

I'm not sure I have a “server with domain control”; I've got a bunch of Macs and PCs with static IPs, access to network folders on a NAS, consistent naming, though. I make good use of $JOB and $HSITE so it doesn't matter which machine I'm working on, they can all access the same assets (OTLs / maps etc) the same way.

Can't quite get my head around why, on a heterogenous network (Macs and PCs), they need to be set up to run Houdini from a shared folder as opposed to a nice speedy local SSD, though. I'm guessing it's down to file system differences. You'd think each client would be able to use its own copy of Houdini to service jobs.

I'll have another go. Have to be a bit careful I don't sink too much time into it, gotta actually do some work too. Thanks for your advice

If nothing else, this conversation has forced me to think this whole thing through: the tools I'm used to in AE are fast and efficient, and give you that 3D-obscured-with-depth thing, but they are overly simplistic. There's no way to light (well, shadow) your elements sensibly. If you want anything approaching photorealism, DCMs'll get you halfway there, but really you just need to have the geometry there in the scene.

That said, it's a valuable tool for motion graphics. I'll stick it in an RFE and see what they say

But! new challenge for me.. there is points with 5 edges and others with 6 edges… now just trying to find a way to scatter one object only if the point has 5 or 6 vertex under it… lets see if I can get far with that.

Look at the Group by expression SOP in my file: it adds all the points that have 4 connections to them to the group, using the handy pointprims function (which returns an array of any primitives making use of the current point). Change that 4 to a 5. Now you can use that group in a Copy to points SOP to only copy your mesh to one set of points. Duplicate the Copy SOP, stick an exclamation mark in front of the group name - !group1 - to copy only to the points that aren't in the group. Merge the results

Hehe - it's just struck me: even if I can't get the geometry in, I could possibly create it on the fly using the depth map and displacement on a grid stuck to the camera. It'd only work for a simplistic effect (it's not like you could light it sensibly, nor get shadows) but at least it'd be 3D.

But that is getting a bit silly.

Shame there ain't a built-in option - it's such a simple thing to implement, programmatically - but thanks for your advice

> why not occlude your clouds/fog render in comp stage instead of rendering it with the provided ZDepth pass?

That's a different effect though - while it'll work fine for homogenous fog, the moment you have any 3D detail in your clouds, and a moving camera, simple occlusion / luma-matting doesn't work any more. The perspective'll break - while a still may look fine, objects in the middle distance will have elements of the furthest cloud appearing over them. Move the camera and it'll look wrong.

Trouble is, I can do this simply enough with other tools, but I want to use Houdini because of course I do.

It may not be possible: Houdini is big boyz toys, and I dare say the ethos is “go DCM or go home”, but I've gotta ask.

Quick example of how it works in other packages: I have a render -



And a z-depth pass:



Trapcode Particular, along with many other After Effects plugins, renders elements in 3D space. And it can use an existing depth pass (if you have one) to occlude any objects it's rendering if their depth would put them behind the Z-depth at that pixel. So using that z-depth pass, and rendering lots of little “cloud” particles gives me this:



Which then comps perfectly over the background plate:



OK, so it's not an attractive example, but you see what I mean: this is not just a depth-fog effect, it actually occludes 3D elements properly, based on the Z-depth. Which means as the camera moves forward, the clouds' perspective and 3D-ness stays intact, and stay locked correctly around the buildings. *

Can't do that with a luma matte.

In an ideal world, I'd have the whole scene's geometry within Houdini with a holdout shader; as it is, I'm asking Mantra to pretend there's holdout geometry present at such and such a depth for this pixel etc etc.

If there isn't a super-easy “oh yes, just plug your depth map in here”, it'll still be possible, just super difficult (Deep Cam stuff).

¯\_(ツ)_/¯

* part of the reason I want to use H to do this instead of the Particular approach above is that Particular's rendering those clouds as cards, so you do get that nasty “Doom” style fog layers cut off sharply by angled surfaces, which you can see if you look closely at the nearer buildings. Still better than nothing though

I have a render plus a depth map created in other software - where / how do I plug it into Mantra (and set the black/white distances) so it knows when to occlude stuff? I need to add clouds/fog to an existing render.

Sorry, I'm sure this is ultra-basic, but googling only seems to bring up how to create depth maps, not how to use them within Houdini.

It's just plain ole' Z-depth (EXR sequence), not a DCM.

… and right now less clicks = more likely to get home on time. But I promise to learn the constraint approach. Thanks!

Oh my goodness. That's wonderful. Thanks a million

And no, I wouldn't have liked to type this by hand (let alone copy and paste and tweak for Y and Z):


Crikey - I didn't know you could stuff that much code in one of those tiny little boxes either

I don't think RS supports anything other than density for volumes - there doesn't seem to be any way to use textures. (boo) … I'd love to hear if I'm wrong

You can use a texture map instead, if that suits, and use a COP net to generate the mask.

Houdini can handle masses of points, though, so it may be worth just using a high res mesh - plus it lets you add things like the Paint SOP to tweak density after the VOP if you need it.

Alternatively look into the Heightfield tools..? There's an HF Scatter sop precisely for terrain use

Newb here, too. Pleased to meet you

Not quite clear on why you want to scatter - Scatter op is usually when you want randomness to the points' positions.

One issue you'll have is that in a poly (geodesic) sphere the edges aren't all the same length.

Had a quick play, though; definitely not there yet, but there might be some clues in there. Note the use of Poly Frame, with the Tangent attribute changed to N - this gives you normals that are actually tangential. Also the use of primpoints in the Group Expression to find all the points connected to only 4 others (which happen to be the structure's midpoints)

There are issues (rotation of some of the tubes is wrong) but I ran out of time.

Suspect if you're trying to create a tent you may be best off with a poly wire approach, then add stuff to the points after. (Note the weirdness: to delete all the primitives in a bit of geometry, but leave the points, you need to use the Add SOP. That's Houdini for you