Ocean Cliff Waves

January 27, 2021 Bradley Sakaguchi

wave velocity VOP net. multiplies the velocity field by a vertical gradient controlled by a ramp parameter. eg. if the wave goes above the initial surface of the water I can make the vertical ramp 0 above that point and multiply the velocity field b… — wave velocity VOP net. multiplies the velocity field by a vertical gradient controlled by a ramp parameter. eg. if the wave goes above the initial surface of the water I can make the vertical ramp 0 above that point and multiply the velocity field by that ramp.

The curl of the waves were made by using the ocean waves SOP node. The problem with it is that it wouldn’t curl like the waves I was trying to achieve. I fixed this by using a VOP net in conjunction with the slope of the ground which in real life causes waves to curl.

fun fact: a gradual slope causes a small slow curl (like mine) where as a sharp elevation change in the ground will cause a more drastic immediate curl.

example of the velocity field for a wave. Lots of velocity below the surface and and almost none above. The big wave velocity is the yellow/red part.

This is the wet map it was generated from the particles of the flip simulation. This was generated using the vex code below inside of a custom solver (further explained below). The data was then sent to a COP network where it could be used as a texture in the material network shown below. This map was used to increase the secular, decrease the roughness, and darken the base color.

material network for cliffs with wet map.

This is the vex code that generates the wet map. It is in volume wrangle meaning that it will loop through each voxel of the input volume, the volume being a 512x512 2D volume which is basically an image. Meaning this code is being executed for each… — This is the vex code that generates the wet map. It is in volume wrangle meaning that it will loop through each voxel of the input volume, the volume being a 512x512 2D volume which is basically an image. Meaning this code is being executed for each “pixel”.
the first thing it does is ask for the closest primitive of the cliff geometry to the fist voxel of the volume and get its closest uv position.
then convert the closest uv position (previously found) to a interpolated position in world pace.
using that position find the closest point from the flip simulation and measure how far away it is
then remap the distance to 0-1 and apply that to the volume as a density value.

Here is some reference footage from my home state of Hawaii all around Oahu

Houdini C++ HDK

November 19, 2020 Bradley Sakaguchi

This a first attempt at making a SOP node using Houdini’s C++ HDK. As shown in the video this creates a select number of points and offsets them by a fixed amount. not the most useful thing, but a good introduction to the libraries, and a nice refresher of C++.

Above is the header file for my “Point_HDK“ class. all the declaration of variables and functions appears here.

To the left is the implementation of the node containing functions for registering the node, creating the sop parameters, and the functionality while cooking the sop.

Ole #5

April 27, 2020 Bradley Sakaguchi

I turned the cig smoke into a volume from the lines from last time and I also added a shader to more properly evaluate the shape of the cig smoke. Right now I think the noise of the initial pyro sim, that drives the secondary cig smoke sim, is a bit to consistent. I may also add an additional advection step to the cig smoke volume after the line have been rasterize to try and smooth out the shapes and give greater variation in density.

Ole #4

April 22, 2020 Bradley Sakaguchi

This is the cigarette smoke for shot 1, but only the lines of the smoke, (non-rasterized). The animation is inside the car moving very fast meaning that the cigarette smoke would leave a long trail. Where in reality that would not happen. If the windows were closed the smoke would act as if it were in a static room, (disregarding acceleration forces which are not in the scene).

In order to counter the car moving I am taking the position of the matadors overall translation controller and subtracting it from the cigarette’s position. This leaves it “static” in comparison to the car this makes the simulation a lot easier and smaller. I will then have to add the position of the car back to the cigarette and the smoke after the simulation is finished.

Ole #3

April 14, 2020 Bradley Sakaguchi

This is the updated lava with a mesh light that helps illuminate it more and decreases the amount of noise as compared to plain emission. I also shaded the sparks in a similar manner to the lava using the temperature to drive a color ramp.

this is the arnold vop net that is located in the arnold light. Its setup is similar to the lava shader but controls the light instead of emission

this s the shader network for the sparks. They do not emit much light so it is just a shader with emission. it takes the temperature attribute of the sparks and drives a ramp that controls the color.

this is the portion of the network that effects the lava. you can see the shop net that hold the shaders for the lava and sparks. You can also see the mesh light for the lava

Ole #2

April 14, 2020 Bradley Sakaguchi

this is a version of the lava that is rendered in arnold. the shader works by using the temperature atrribute on the lava to drive 2 ramps one of which drives the color of the emmision and the other drivex the amount of emmision. This makes the lava get darker as it cools. This is the network for the shader

Ole #1

April 6, 2020 Bradley Sakaguchi

this is the newest render of shot 9.

Car Commercial #9

March 11, 2020 Bradley Sakaguchi

This is the final network for shot 1 consisting of source Geo, the pebble sim, 2 particle dust sims, and a pyro sim

This is the full network view for shot 2 consisting of the cache of the asphalt sim Felipe made. Then taking that cache and creating an appropriate dirt geometry and force for the sim.

This is the DOP net work for the grains. I am using a “popadvectbyvolume” node to get the forces correct and act like the UFO is lifting it. bothe sop solver and that pop wrangle have the code

if(@P.y < 0.01){
@v.x *= 0;
@v.z *= 0;
}

Which stops the grains under the asphalt from moving.

I then take the result and use “fluid surface” to make the points into geometry.

this is the result of the dirt Geo which acts as the source for the dirt. The grey pieces are under the asphalt pieces that move. I then find the normals that are facing downwards with this code,

if(@N.y < -.999){
setprimgroup(0,"down",@primnum,1,"set");
}

then extrude them downwards. I find the connected pieces and also find the bounding box of all the pieces. I turn both into a volume and I dilate the pieces to make them larger and use the bounding box as a mask so it can only expand on the x and z axis.

This is the force that I am using to advect the grains for the dirt. I make a line and set the top and bottom point to different pscales to make a cone. I then fill it with “points from volume“ node and merge it with the original top point. I sort them by Y position so the top point is always point number 0.

to make the actual forces I use a point wrangle, called “set_distance“ and give each point a distance attribute, which is the distance that each point is away from the original top point.

vector goal = point(0,"P",0);
f@dist = distance(goal,@P);

I then have a detail wrangle that loops through each point and over writes the “lrgDist” variable if it finds a point that has a larger distance than the last one in the loop, if it is smaller an “lrgDist“ stays the same.

float lrgDist = 0.0;
for(int i = 0;i < @numpt; i++){
float pntdist = point(0,"dist",i);
if(pntdist > lrgDist){
lrgDist = pntdist;
}
}
f@maxDist = lrgDist;

lastly to make the actual force attribute I start by making a gradient based on the point’s distance to the top point. The gradient goes from 1 at the bottom to 4 at the top. I make this by fitting the distance which range of numbers goes from 0 to the max distance I found in the last point wrangle, to a new range of 4-1. I then find the vector between the top most point and all the other point by subtracting the position of each point by the goal point’s position and normalizing it. I then multiply the force variable by the accel variable and that creates a force that is directional and has acceleration

vector goal = point(0,"P",0);
float accel = fit(@dist,0,detail(0,"maxDist",0),4,1);
if(@ptnum != 0){
v@force = goal - @P;
}
else{
@force = 0;
}
@force = normalize(@force)*accel;

Car Commercial #8

March 1, 2020 Bradley Sakaguchi

Ive been getting new iterations of the particle sim based on the feed back I got from the last session. The two main things is the amount of density that is blocking the view of the ground and a force parallel to the ground that pushes particles from under the car to the camera.

the setup hasn’t changed too much but in order to achieve some more force latterally I added a pop attract under the car

I have also been working on a the dirt under the road for shot 2. This should give us the final layer between ballistic and continuous effects. I made demo of this early in the quarter and it is the same setup of RBDs with grains being sourced off them.

Car Commercial #7

February 17, 2020 Bradley Sakaguchi

This is the most recent test render of new particles. They are rendered as volumes in arnold and motion blur to make them smother but still have texture.

Car Commercial #6

February 7, 2020 Bradley Sakaguchi

Pebbles

This is the new pebble sim with varying sizes. The image to the left is the point attributes I set on the points. I use the @rand attribute to set the pscale.

I keep the @rand attribute on the sim geometry and then use a sop solver in the dop netwrok to then set the mass using the same rand attribute, meaning smaller the pscale smaller the mass.

this is the vex for the mass attribute :

@mass = fit(@rand,0,1,.25,1);

Pyro

This is the dop network for the pyro. It is very simple right now as I fell like the tractor beam forces would be fairly smooth for objects so light.I could also have a similar setup to the pebbles and add a ramp and noise to the gas wind which, if I fell it is lacking detail. If I decide to up the resolution of the sim I may add some more higher frequency turbulence within the sim as the force going up is fairly clean right now, as you can see in the video above..

The video above is the first render tests but I really did not like the result, the lift was far to consistent at the start and the there is no noise in the sim. The screen shots below are a new sim that I think I like better more noise and less consistent upwards force.

Car Commercial #5

February 2, 2020 Bradley Sakaguchi

Shot 1

This Week I primarily worked on shot 1. The shot consists of three main sims RBD pebbles, particles for larger dust, and Pyro for small aerosolized dust. This week I was able to get the timing down for the pebbles and particles, however the Pyro still needs work.

This is the DOP Network for the pebbles. The first SOP solver gives the pebbles a random mass

This is the vex code:
@mass = fit(rand(@ptnum),0,1,.25,1);

just assigns a random 0-1 value based on point number as a seed, then remaps it to mass values that makes the sim act the way I want.

The second sop solve just stops thing bullet objects from continually rolling

The POP wind (parameters shown above) is what gives the pebbles its motion. It uses this HScript expression:

if($F<1,0,(($FF+15)*.07) - fit(rand($FF),0,1,0,11))

everything behind the subtraction sign is to ramp the force up as the time increases, everything in front of the subtraction sign adding noise to the ramp.

The POP simulation is essentially the same setup but with some of the values tweaked.

Here are 2 very rough pyro sims to show off the possible shapes and motion of the dust

another thing that I worked on this week was simulated grass cards. Because we chose to shoot on a day that was not windy and there should be a giant UFO in the sky we need the background plate to have alot more motion in the long grass.

Grass

This is the setup for the grass sim. I copy stamp lines onto a grid of scattered points and randomize their length. Resample to make sure each blade has the same density of points because the more segments a line has the more floppy it is with vellum. I use a pin constraint on the bottom two points of each strand, then hair constraint and crank the bend stiffness. This is then fed into the DOP net with a vellum solver and a pop wind.

.ASS files

There are some more .ass tests, 1st is showing that you can also assign shaders in Maya, 2nd is showing that you can change the radius size of the particles in Maya, 3rd is showing that you can subdivide and smooth geometry. I was also able to add displacement to the geometry but I have not gotten a render of that as of yet. This is by far the most promising method we have found for moving geometry from Houdini to Maya and all the effects for the newest render were done using .ass files on the render farm.

Car Commercial #4

January 28, 2020 Bradley Sakaguchi

These are two still frame renders, the top is Arnold for Houdini and the bottom is Arnold for Maya. I was able to transfer the geometry, lights, basic sharers, and particles (which I have had issues with in the past) to Maya from Houdini.

The next tests I am planning on running are camera, more advanced shaders, volumes, and rending .ass files on the farm.

The above image, is a test of transferring “Cd” data from Houdini to Maya through a Arnold shader.

add Arnold properties to geometry network

This is how to export the .ass files from Houdini

This is how to import the .ass files into maya.

I was also looking for more reference from movies to get the timing and choreography of the UFO tractor beam.

Car Commercial #3

January 26, 2020 Bradley Sakaguchi

This is newest dirt sim. I worked primarily on figuring out a method to make the RBD sim a higher resolution. During meeting I was given the feedback to look into the new Houdini 18 guided sims which I implemented into this iteration of the sim. Not only did this make the sim significantly faster but it add a lot of the complexity that I wanted, particularly with how the pieces start as large compacted pieces of dirt and break into smaller chunks. The higher resolution rbd and increased particle count also gives the impression of flow in the dirt.

The next improvements I plan on making are fixing the uniformity of the sim adding a a gradient to the sim making the top layer of the dirt “finer” and have the guides effect it less, while the bottom layer a “rockier” and more heavily effected by the by the guide.

This is the setup of the guided sim

input 1 - high res fractured geometry to be guided

input 4 - collision objects

input 5 - low res guide sim

I tweaked the strength of the guides on the high res sim as they were a bit strong to start.

More Reference

tldr Examination and Description of Soil Profiles

Car Commercial #2

January 20, 2020 Bradley Sakaguchi

These are the dirt sims we have so far. The basic set up behind it is a fractured rectangle held together by hard constraints that is being used to source vellum grains. I will also be adding pyro to this for the small aerosolized dust.

The biggest issue that I ran into was the fact that the dirt RBD pieces need to be so small to look correct, but Bullet doesn’t like small geometry. I just ended to scaling up the cube pieces and increasing the “Time Scale“, which still needs to be tweaked as it looks too fast currently. I may try to use RBD as apposed to Bullet but this will be for later R&D.

These are the 2 DOP networks (left = RBD, Right = Vellum Grains )

I added a little sop solver to the post solve of the rigid body solver to stop the little bullet pieces from rolling away. I plan on adding a lot more different types of constraints and dialing in the ones i have now as they look a little spongy currently.

I am using the “debris source” node to spawn the grains from the RBD sim. In the DOP net I am using continuous spawning “Grains_source_vellum“ node. I had to dive into that node to add a point wrangle that reads the age attribute of the source points created by the “debris source” node and if they exceed a certain age they are deleted.

in the remove broken sop solver (in the RBD DOP) I added a primitive wrangle that sets constraints to the “broken” group if the force or distance attribute exceeds a certain threshold. This allows the hard constraints to break

Car Commercial #1

January 12, 2020 Bradley Sakaguchi

Soil FX reference

These are some of the videos that we are going to be working from when creating the soil FX. it is going to be a both asphalt and soil so making sure that we get a realistic version of both of these materials is going to be key.

This video is a great example of the characteristics of asphalt. the road is a combination of rocks and bitumen which is a substance that seems like a solid but is actually a highly viscous fluid. This combination makes it flexible but also tearable. I think this effect can be achieved through a RBD sim with hard constraints creating a “soft-body“ looking sim.

pitch drop video:

https://www.youtube.com/watch?v=SkpZAg4RkGI&t=24s

this is a great soil sim Peter Sanitra! There is a lvl 80 post out lining the general unspecific process he used and we will more than likely use this as a guide for the soil.

the general idea behind the soil sim is to make a RBD sim with hard constraints that are actually soft and allow some flex. Then create a Pop or Vellum grains sim that is sourced from when the RDB constraints break, and same for the pyro sim for the aerosolized dirt.

https://80.lv/articles/making-tree-soil-animation-in-houdini/

Candle #3

May 23, 2019 Bradley Sakaguchi

Candle #2

May 16, 2019 Bradley Sakaguchi

This is the DOP network. I have added some disturbance to the Pyro and shaped it a bit more, although some more does need to bee done. I have primarily been working on getting the wax to behave correctly. The “gas temperature update” node’s temperature diffusion was not giving me the result i wanted so i made a new temperature diffusion network. I also shaped the temperature field from the Pyro to give me better results.

Temperature Diffusion Method

in the data stream add scalar field “waxtemp“ that is the same size as the Pyro container.
copy flame field to particle attribute “flametemp“
copy “flametemp” to new “temptemp“ field
apply particle “temperature” attribute to “waxtemp”
blur “waxtemp“ field
gas linear combination - “waxtemp“ * .8 + “temptemp” * 4 = new “waxtemp“
geometry wrangle - @temperature += @waxtemp

This is a little sub network i created to visualize fields created in Dops. You can type in a field name in the parameter menu of the sub network and it will change which field it visualizes. Small but very helpful to understand what is going on.

Candle #1

May 9, 2019 Bradley Sakaguchi

This is the current candle sim it has all the functionality needed right now. I am in the process of tweaking and refining the sim to look like an actual candle.

This is the candle model. I started with a line and UV’ed it top to bottom I can then use the UV attribute as the position of the ramp and the set value at that position is the “pscale”. I then used “polywire” to create the geometry.

This is the code for the wick it, creates high fuel and temperature attributes that travels down the length of the wick. It starts by dividing the total number of frames by the current frame number, thereby returning a value between 0-1 ( there is a… — This is the code for the wick it, creates high fuel and temperature attributes that travels down the length of the wick. It starts by dividing the total number of frames by the current frame number, thereby returning a value between 0-1 ( there is also a multiplier to slow down or speed up the growth). I then fit that 0-1 frame value to the total number of points on the line. I the explicitly cast this to a integer to drop the decimal value and this gives a point number that travels downward like a wick burning out.

This is the DOP network for the candle. The challenge of this sim was to get the flip fluid’s viscosity to be effected by the temperature field of the pyro sim. The “gas temperature update” node makes the temperature of the fluid particles to effect… — This is the DOP network for the candle. The challenge of this sim was to get the flip fluid’s viscosity to be effected by the temperature field of the pyro sim. The “gas temperature update” node makes the temperature of the fluid particles to effect its viscosity and also allows the temperature to cool over time and diffuse through the fluid. To transfer the temperature field of the pyro to the flip object I use the fetch data node and then apply it to the flip object using the apply data node. this gives me the temperature field but i need it transferred on to the particles so I use the “gas field to particle” node to transfer it. That’s how to transfer the temperature data within the DOP network.

Vine Gen #3

March 13, 2019 Bradley Sakaguchi

Vine Code

Scatter points on a user defined mesh

initialize a root point with a random direction vector
types of points (point groups)
- root points: will search for “nodes” to turn into roots, they will “die” the next frame (points that are “dead” can not grow and can not be grown to)
- node points: valid points to grow to
- stem points: a collection of all root points that have died
the root node look for a specified number of points in a specified radius
- (the close points are put into an array)
loop through the close points
- draw a vector between the root and all close points
- compare the created vector to the direction vector on the current root
- find the close point that has the vector that most closely matches the direction of the direction vector, that point will be the new root node
- kill the current root and place it in the stem group node and apply the drawn vector to the new root as its direction vector
create a poly line and assign it 2 vertices the first is the old root the second is the new root
We now have a new root node with a new direction vector and this process is repeated every frame

Leaf Code

In the Vine Vex create a new “leaf root” at each frame positioned at the normal, if a probability is true (place new leaf root into leaf group)

Create an array of every point in the leaf group and loop through it
- Get position and normal of leaf root
- Subtract a gravity vector from the normal at each frame
- Create a new point at the edited normal + position
- Set a “leaf id” attribute to keep track of how many points have been generated per stem
- Also transfer edited normal on to the created point to get an iterativly add the gravity vector
- Add primitive just like vine
- Set primitive to new group for skinning
Check the “leaf id “ at start of for-each and if it is larger then a user defined range of numbers than stop the current loop there and continue to the next loop