Yuki Murakami 2024 at RSP

I did not have much except for the passion in making FX when I started this journey. So finalising this project makes me so proud and honoured. It is thrilling to share this work on Rookies.  I hope this blog will be helpful for an aspiring FX artist like me , who is scared of Houdini, in some way for their own project.  I own this project to the RSP education team mentioned at the end.

The Dune

In this project, I aim to achieve realistic look that is visually appealing. 

Ideas & Reference

I started by looking for images of dunes at sunset to visualise the look I was after. Additionally, I looked for reference videos of cars driving on dunes to see how the sand and dust interact with the car. It was interesting to see that they behave like a liquid.

Rigging & Animation

For the jeep, I used a "1987 Jeep Wrangler".  This model has a high level of detail. The only downside is that it has 3,800,000 polygons with 8k textures on each component. It was challenging to reduce the polygon count and optimize it in Houdini. I used the Rigacar addon to rig and animate the car in Blender.

Modelling & Texturing

I used the height-field node to create the dune-like geometry. More specifically, I added the streaky look to the height-field by using the vein node in VOP. Then, it was brought into Substance Painter for texturing.

Simulation

Obviously, simulation is the most crucial component of this project. By comparing it with the reference, I worked to fit the simulation into the shot. There are two dust simulations, one particle simulation, and three layers of ambient dust simulations.

Firstly, I set up a particle operator to emit particles from the wheels (1). These particles are set to only last a few frames. I used this setup as the source volume of the pyro simulation (2). The velocity of the wheels is imported into the pyro sim (3). Lastly, I used this dust sim to drive another particle operator to realize a sandy, grainy effect (4).

One mistake I made was to separate the pyro sim for each wheel so that I could control them individually with smaller disk space. However, they didn't seem to be coherent with one another. As a solution, I imported those sims into one big dust sim as a velocity field.

There are two useful techniques to point out. One is to use "volume sample" to remove particles in relation to a certain object. The volume sample function gives the distance between the particle and the object surface. We can use this distance as a threshold to remove particles (1). As for the velocity field of the wheel, I added inward velocity in addition to the rotation speed of the wheels. This enables me to blend the velocity to control how much the dust is pushed away from the wheels (2).

Composition

With all the assets rendered in Arnold, they were imported into Nuke. I made sure the colors looked coherent with each other, and no object stood out. Minor adjustments were made using roto paint.

The Bonfire

This project aims to replicate real-life references. I developed my understanding of pyro simulation to the next level through this shot because I carefully adjusted the settings and parameters one by one to see what each parameter was doing.

Approach

I started by focusing on just the fire in the reference: how fast it rises, how much its size deforms, how noisy its size deformation is, and how fast the noise moves. These correspond with the scale of the temperature attribute, buoyancy, swirl size, grain, and pulse length. This time, I only utilized turbulence for deformation, which allowed me to see exactly how a parameter affects the shape of the fire. The shutter speed of the camera and the substeps of the simulation also have a great impact on the shape.

Then, the look of the pyro is refined by the pyro bake volume. Settings on turbulence and pyro bake volume are the most crucial nodes to adjust. Masking, source range, fire intensity ramp, and fire color ramp are attentively experimented with. The images below show how different they can make it.

Regarding the smoke, I copied the pyro solver for smoke and adjusted the grain of turbulence to make the smoke more fluffy and rough.

One of the challenges was the voxel dots of fire on the wood (1). To remove these dots, I converted the wood geometry to VDB and subtracted the flame VDB of the bonfire from this wood VDB, clamping the value so it does not exceed zero (2). This successfully removed some of  the dots on the wood (3).

After color experimentation in the pyro bake volume, it was rendered in Arnold. Then, I graded the color of the fire to match the reference as much as possible while ensuring no 3D assets stood out.

RSP team

Lastly, this project could not be achieved without the support of RSP education team and members. 
Especially, Thomas James: Effect Artist/ Instructor, Dan wills: Effect Technical Director/ Instructor,  Micahel Kranitis: Teaching Assistant and Masahiro Yanagiya: Senior Lighting Artist. Thank you so much:)