I’ve been having a serious look at using spheres and also boxes to represent the collision
shape in bullet simulations. Spheres are great in that you can rely on the geometry to not get caught and hang on to each other. Boxes offer a nicer settling motion in certain setups too. The main reason I wanted to look at it, was to avoid always needing to use glue constraints. Sometimes I want a nice quick voronoi fracture that uses the clustering option, and this yields shitty collision approximations with convex hulls.

Enter vdb sphere packing! It allows me to fill the wacky geometry shapes and use those as colliders. I just save out my points, and transform the original pieces. Voila!


Dot Product & Normalize

September 25, 2018

These are two of the most used functions in Computer Graphics, and quite a lot of the time, the underlying process of how it works, is ignored.

Normalizing a vector is essential for a variety of reasons, and below is the long way around to do it, as opposed to the normalize VOP.


We need to find the length of the Vector, as this will be used as the divisor for the x,y,z components of the Vector to be normalized. This is done by taking the x,y,z components and taking each to ^2 power. Then we add those values together and find the square root. Now we have the length of the Vector, and  can use it as the divisor. Voila!

Now it’s time to do the Dot Product! Mysterious I know. But very handy in shading/lighting and other areas of Computer Graphics. The Dot Product is the Cosine of two normalized vectors. Cosine being the trig function equal to the ratio of the side adjacent to an acute angle (in a right-angled triangle) to the hypotenuse. That’s pretty neat!

Below is a quick animation of the Cosine around a unit circle. The unit circle being a radius of 1 of course. The Dot product of two normalized vectors, which we will see is just multiplying their components and adding them together, creates a trigonometric function.

Here is the long way of doing it, instead of using the dot product VOP.


First we normalize both Vectors. Then we split the x,y,z components of each vector and multiply them. Adding the results of this all together. Bingo! Very Simple really isn’t it. Additionally, using a trig VOP to use arc cosine along with a radians to degrees, will give you the actual angle between those two vectors.


I Have been working on this for the last six months.

2018 Reel

July 2, 2018

Blah blah, doesn’t have the awesome work from current projects, but it’s not too bad.

Tinkering with using uniform or varying volume, along with reflection and refraction to create more interesting attenuation of light.


So a major pain in the arse in VFX land is trying to get a particle sim, well any sim that is stored as points, out of Houdini into Maya for rendering. Yes yes we’d all like to just use Mantra, but a big part of VFX lighting is done in Maya.

Autodesk’s Alembic tools in Maya are shit. End of discussion. Nothing like the delayed load, deferred to the GPU, streamed off disk benefits we enjoy in Houdini. The J CUBE guys in Japan have been making a product to solve this cross-walk. Multi-Verse is it. And by it, I mean IT. Does exactly what it says on the tin. Lightning fast loading of Alembic or USD into Maya. A typical 10 million vert model will load in under 1 second, there in your viewport, to tumble around with no lag. You have a choice of geo, points, bounding box, or hidden for display, and the ability to actually scrub your viewport with amazing feedback. It deals with packed geo from Houdini automatically, so your Alembics out of Houdini, if for example, are instancing a few bits of unique geo to a bazillion points, then your packing before instancing will be respected. That means tiny Alembic files due to the multi-verse procedural being able to feed to Arnold, prman, 3Delight in a manner you’re used to in Mantra. Full support for Mantra style point velocity blur as well.

What I have never been happy with, is how hard it is to simply save points, with attributes, and have them render with a lightweight primitive, that can utilize those stored attributes in Maya. What follows is a nice little walk through of the process for getting a particle sim out of H into Maya, and rendering with 3Delight.

Let’s make a simple popsim, and set a couple attrs for pscale and color!


Alembic ROP settings!

So we are using Houdini 16.5.323, as from 16.5 onwards per frame Alembic caching is sweet! Much better than a huge single Alembic. Untick render full range to output a per frame, and enable full bounding box tree, as this makes life easier in Maya. Do not enable motion blur, it’s not needed. We have point velocity in this example, but even with fixed point counts, for regular motion blur, multi-verse will correctly interpolate, even with just one sample(the cached alembic frame).


Our lovely Mantra render!


Loading the Alembic with multi-verse

Use the green loader button, and since we are loading per frame, tick the option and remove your existing frame number, and replace with #. So, dude_1234.abc = dude_#.abc

You have options for full geo, points, bbox, and hidden, on the same shelf.


3Delight assignment panel

If you’ve ever used 3dl in the past, you know this panel well. In the new version of 3dl this will no longer be around, but for the next couple months we have to live with it. Select your alembic that we just brought in, and launch the panel, from the shelf. On the Attribs section, hit the little triangle and it will add a “delightGeoAttribs” panel to your Alembic. This is for doing all many of per object overrides, so much so, that it was one of the major selling points of 3dl.


Delight attribs, and Attribute re-mapping!

on the left is the panel that opens when you hit “Add / Remove Attributes” on your delightGeoAttribs panel that we attached earlier. We only care about the Particles options, so tick All Particle Attributes and close. What you will also see, is there is a Multiverse tab on our delightGeoAttribs. Open the Geometry Mappings and re-map our pscale and color to: pscale=width; Cd=Cs to make everyone happy! Width is the ri spec for radius, and Cs is color. Now when 3Delight renders it will pick up the width correctly and the stored per point color will come through in the diffuse of the 3dl shader. Point velocity motion blur scale can also be adjusted here, to accommodate different scene scales.


Delight particle settings

When we added the Particle Attributes, it gives you the following. Here we make sure that our points are rendering as Spheres, in Object Space, and I have reduced the particle scale globally to closer match Houdini. I’m sure it should be an even number, but meh. Interesting point to note, you can set your points to be in Raster Space, and they will always render the same size, in pixels, no matter their depth. So you can easily mimic the behavior of Krakatoa!


3Delight render pass

3dl has always had a nice pass manager, and all you need to do is to create a pass from the 3Delight menu. Once you do, set your camera and environment light in Scene Elements. The default render engine is the path tracer, so just enable Indirect Illumination and motion blur, and off you go!

It should be noted that the completely re-written 3Delight V13, which is coming soon, has a much more streamlined interface, and you will literally be only changing a couple settings. OSL based, designed for Interactive rendering, it is unbelievably quick, and if you have still got a hangover from the previous version, with it’s plethora of overrides and fiddly TD options, then you will be very pleasantly surprised.


Render time!

So our Alembic points from Houdini render perfectly. Motion blur, color per point and radius per point too. With the new version of 3dl being much quicker, which the current version rendered these in 60% less time than Mantra, coupled with the multi-verse tools doing the cross-walk, it’s an exciting time indeed.



1.) Grab the latest studio pro download for Linux. (12.5.x) as of writing, but  I’m using 13.1.7 alpha.

2) Extract the contents, I’ve just gone “/home/dude/Downloads” and make a new folder in your home directory, “/home/dude/programs” for example.

3) Open a terminal and cd to “/home/dude/Downloads/3delight-13.1.7-Linux-x86_64” and run the following: “./install –prefix home/dude/programs”

4)cd to /home/dude/programs/3delight-13.1.7

5)run: “unset DELIGHT”

6)run: “source .3delight_bash”

7)cd to home

8)sudo gedit .bashrc (and add the following lines)

# set DELIGHT path

export DELIGHT=/home/dude/programs/3delight-13.1.7/Linux-x86_64
source /home/dude/programs/3delight-13.1.7/.3delight_bash

9)I had issues with these env variables not being present, so set these too, with the following command.  “export MAYA_PLUG_IN_PATH=/home/dude/programs/3delight-13.1.7/maya/2017/plug-ins”, and MAYA_SCRIPTS_PATH, DL_DISPLAYS_PATH too.

10) Set the frame buffer to be i-display in 3delight prefs in Maya, before you render, or it will crash Maya.

I launch Maya from a terminal, that way everything is set right. Also, if you plan on adding your own custom OSL shaders, set this env too.