eikonoklastes

Not much to see in the wireframe. Just a standard sphere and a curve projected onto it.

Hello. Another C4D-inspired post, from here. To tackle this, a displacement shader can do an excellent job at render time, using standard geometry, that would otherwise require unreasonably high poly counts, if modeled. The below image, rendered in Mantra, uses a curve (rather than a texture), to define the displacement area, and the shader can use a ramp to create the falloff. Mantra might be slow, but the quality it produces is evident from the extremely clean displacement it generated. I'm not fully comfortable with Karma, but I believe I can reproduce a similar image with it. If you inspect the attached scene, you'll see a bog-standard setup with the geometry. All the magic happens in the material shader, where you can query the distance between two objects (at render time), and then drive a displacement with that information. Displacement Shader with XYZDist (Mantra).hiplc

My knowledge about volumes is pretty limited, but I think a Volume Rasterize can do something close to what you want. scatter_on_volume_density.hipnc

Scatter can read the density values in volumes natively. You don't need to enable the Density Attribute on it - that setting uses the density attribute from points. Just initialise your values in your volume and you should be good to go. Extra note on the Force Total Count parameter on the Scatter node. If you have that enabled, and your density values are too low, Scatter will forcefully add points outside your defined density range. So, either crank up your density values, so that they exceed the Force Total Count value, effectively using that parameter only to prune the point count, or turn off Force Total Count, and let your density values drive your point count.

Quick follow-up. Lit Spheres is also extremely useful in shaping and lighting particle sims. With just the standard view, the particles just look like a generic mass, but with Lit Spheres, you can really bring out the shape and detail, without needing to do a render. Here's are comparative top views:

Hahaha, yeah. Trying to show scale on 10 million points individually is tricky. I cranked it up to demonstrate how well it works even at absurd point counts. I do zoom in on them in the video at around 15 seconds.

The C4D posts are serving as a good source of inspiration. Saw this thread and decided to play with some rigid body dynamics, something I've barely touched in Houdini. mplay_I11tJhYlsV_H.264.mp4 Scene file attached. A 250 frame sim took about a minute on my 5 year old machine. Rolling RBD Sim.hiplc

A pretty common workflow in Houdini is to copy geometry to points. When dealing with a large amount of points, the copy process can be a bit slow, and make the viewport sluggish. A good way to preview the copies and tweak their sizes and colours without actually copying any geometry to the points is to change the display of the points themselves to Lit Spheres in the Viewport Display options (press D with the mouse cursor over the viewport to bring up the options). Here is a preview of 10 million points running effortlessly on my old GTX 1080 Ti:

Yes, as long as you understand their limitations, they're not inherently bad. I have used them extensively over the years on several models, with no detriment to output quality, so I'm not a fan of the general "ngons = bad" sentiment. They're bad if you use them badly, but that's true of pretty much everything. About the ngons that I posted that seemingly mysteriously deformed properly, I probably should have included the wireframes in the screenshot, but I did address it in the post, where I said the DIvide SOP can handle the deformations if needed.

It's a fair point. SideFX has only very recently shown a keen interest in making Houdini more approachable. I started with v17, and things were pretty brutal then - almost no beginner learning material, and a very heavy reliance on VEX to get even very basic stuff done. If their development pace since then (which has been astonishing) can be maintained, hopefully things get a lot better Soon™.

What's this ngon stigma I keep encountering? If it's not deforming, it's time to make ngons great again.

If randomness doesn't float your boat, maybe some smooth transitions instead (play with the settings on the Attribute Adjust Float). And the topology is still all quads of course

Why stop when you're having fun? Here's some procedural randomness thrown in for good measure. Circle Holes without Booleans_MA05.hiplc

Very nicely done. Your setup is leagues simpler than mine. Really like the way you're already exploiting Group Expression. Having said that, you don't get precise control over the number of points in the circles that I do, so I'm calling this one a draw! Regarding the corner sharpness, the Subdivide SOP can read weight attributes on points, vertices and prims. For points, you use cornerweight (this is detailed in the docs). I've attached a scene that also uses another trick to grab only the corner points by measuring their curvature. Circle Holes without Booleans_MA04.hiplc

Been having a bit of fun with the Circle from Edges node, and used it to create this fairly efficient setup. No booleans or subdividing needed, and if it needs deforming, the excellent Divide SOP takes care of business. Scene file attached. Circle Holes without Booleans.hiplc

I'm with @MJV on this one. The documentation is very hit or miss, and searching for "normal" or "move along normal" should return Peak, but it doesn't at all. You either know it, or you don't.

I'm not going to sugar-coat this one. There are many Just Houdini Things™ that you are going to encounter. From inconsistencies in UI, to downright straight baffling workflows in some places. It's what got me to nearly drop the app when I was learning, and return to the warm comfort of C4D. All I can say is suck it up and keep grinding until it all makes sense. Is this the worst possible motivational advice? Possibly yes. Hopefully things fall into place and you can start being productive with the app soon, because the pain pays off, and pays off massively. I spent 10 years with C4D and have become significantly more capable with Houdini in 3, because it just freed me from having to depend on a black box, at the mercy of Maxon's (then) not-amazing development pace.

I think a quick scan through the docs would serve you well, specifically here: Viewing the scene (sidefx.com) Not sure what to tell you about the focal length control. I guess SideFX have never found it to be that important. If it's important to you, drop them an RFE (Request For Enhancement), and explain why you think it's important. They read every single one, and will respond to you if they need more info. As for the FOV, I've never encountered a production where that was ever a factor in building a shot. Can you expand on why you need this number? Please don't get me wrong - I'm not saying that it's not needed, I'm just trying to learn where and how people use it. Meanwhile, perhaps this page is of some use to you - Match Houdini camera lenses to the real world (sidefx.com)

I don't believe Houdini has a built-in shortcut for focal length adjustment. If this is something you adjust frequently, you can drag the Focal Length parameter (by its label) into the viewport and get a HUD slider for it.

There are ways to mitigate the mess, but that's not the correct approach, for sure, and you'll have to get into Python likely to address it. The better way is to import into SOPs, and use the name attribute to select and process your pieces. It does require a bit of research and a unique approach, but Houdini is extremely efficient at dealing with pieces via attributes.

The scaling (pscale) and rotation (orient, N, rot) are attributes that can be added to the points, that can be exploited down the line by certain nodes (like Copy to Points). The points themselves aren't being rotated or scaled. Keeping with the Copy to Points example, the copies get their positions from the points. Where do the points get their positions from? From wherever you want. Sticking them to a surface is one way to deform them. From that situation, you have described one way of moving the points (along the normals of the scatter surface). What if that's not what is needed though? What if the positions just need to be jittered a bit? Or slid along the surface? Or move along a separately defined vector? Rather than account for every possible way to move the points and build that into a bloated Scatter node, you can use any tool post-Scatter to do that that can deform the points (or pre-Scatter to deform the surface that you are scattering on). That's the reason we didn't mention a solution to transform the points before, because you didn't mention how the points are supposed to move.

Why is it a workaround? The Scatter node is designed specifically to scatter points on a surface. Otherwise you're just creating points in space, and there are other ways to do that if you want.

The points are being scattered on the surface you choose. To translate them, deform the surface itself, and the points will follow.

If your asset has a bunch of sub-objects, if you bring them in via the standard FBX import, you're going to have a bad time, because you'll end up with a million nodes that are impossible to work with effectively. The better way is to bring the object in via SOPs (as a single File node) and exploit the name attribute to isolate and edit your objects. If you can share a model, or a part of a model, we can look at how best to work with it.

Here is the result I got, and it being Houdini, of course you have to add the extras like variable thickness in the middle and variable radii for the pillars. It's also entirely procedural, so all these can be changed on the fly - Number of pillars Thickness range for the middle The entire shape of the pillar is driven by a ramp, so it can be adjusted for different looks Height of the ceiling (everything will stretch to adapt) Density of the pillar distribution Width and depth of the area The scene file is also attached. I'm happy to field questions about how it's built, or preferably, see someone do a better version. Organic Shape.hiplc