5
Mathieu KERAUDRAN

Mathieu KERAUDRAN

NantesMember since May 2025

I would like to become a FX Lead and working on challenging productions.
In 2009, when my parents took me for Avatar 3D in theatre. I was amazed by the movie and the realism. It was decided I wanted to work for the industry to create my own universe, my own images.

FXSoftware & Pipeline Development
Editor's Pick

Reel Breakdown

Hand-selected by the VFX Engine team

FX
Recent Graduate·ARTFX·September 2025

## Two Kings - Arrival of the golem & Hand detach

For this shot, I was responsible of all the FX and the crowd.

The golem is made of almost 400 parts. Each of them are fractured and simulated to create an organic look. After that, I had to create a more cohesive look so I added peebles around the big rocks using velocity fields. I added falling rocks and peebles, smoke simulations.

For the crowd we used mocap to save time but I have to create a lot of simless transitions betweens cleaned mocap animations.

## Two Kings - Creation of the golem

Theses shots were very challenging. I had to create a magic RBD FX to create the golem. To do so, I used a lot of different sim layers to create a cohesive creation. I used the same layers I told you before but I also had to merge the debris from a big facace (simulated), the generation of little peebles going to the golem and a grain simulation to merge everything together.

## Le Routard - MPC

I did an intership as FX TD at MPC Paris & Liège.

I created a jacuzzi smoke for few shots.

I worked on a film with a full sequence in Himalaya. So I had to create blizzard, snow, and cold breath.

## Dunkirk match reference

Here’s a project I completed 2 years ago. As a fan of the movie Dunkirk, I decided to recreate one of its iconic shots. My goal was to learn the Axiom Solver from Theory Accelerated. The project was created using SideFX Software's Houdini. It was a great exercise to match the reference footage. While it's not perfect, I learned a lot !

### Process:

I sourced a 3D model of a Spitfire online. The mesh had several issues, so I worked on cleaning it up. I used blasts, Voronoi fractures, and numerous transformations to match the reference.

Afterward, I identified three main sources of fire by studying the reference footage: the right wing, a small burning object on the ground, and the primary fire on the right side of the Spitfire. The setup is similar for all three.

First, I painted a mask attribute and added attribute noise using VOPs. I then used this mask to multiply source attributes like density, pressure, and temperature. I added different-sized noises and merged them.

Next, I created spheres as the center of each fire source. This is the area where the fire is larger, hotter, and carries more pressure. I transferred attributes by distance and multiplied my source attributes.

After sourcing the fire, I rasterized my points to create my emitter and simulated the fire layer in Axiom.

I then created the smoke layer by merging the same emitter as the fire and incorporating the fire velocity field.

To generate embers, I used the advect by volume node to capture the velocity fields from the fire and smoke.

For the debris, I emitted smoke points into the fire emission area and established collision behaviors such as sliding or bouncing.

This project was set up in Solaris, rendered with Karma CPU, and composited in Nuke.

## City Explosion

Here is a project I did also 2 years ago. I practiced layering with my main explosion setup in Houdini by adding a blast, a shockwave, trails, sparks, and debris. I also created a procedural destructible building HDA to enhance the render. Additionally, I created a window destruction simulation to further integrate the explosion.

### Process:

I started by creating the layout. My goal was to give the impression of being in a bombed city from the perspective of a 4x4 vehicle. To achieve this, I developed a Houdini digital asset for a procedural destructible building.

First, I scattered some points and copied boxes onto them. Then, I used poly extrude to bring them to the ground level. After that, I divided them by floors and created tools for inner walls, windows, and other structural elements.

All the buildings had to be destructible, so I was careful to avoid any interpenetration of elements. Then, I wedged different buildings and fractured them using material fracture. I used constraint rules to create connections between the different parts of each building.

To destroy them, I used a foreach loop to create metaballs based on a scatter. These metaballs were injected as velocity forces in the RBD solver to procedurally destroy the buildings. To post-process the buildings, I used rbddisconnect faces to prevent shader issues and added my material attributes.

After that, I created the ground, which was full of craters.

Time to explode things... Wait, why? Initially, I wanted to create an explosion, but it seemed odd to explode nothing. So, I decided to create a large smoke plume with fire to justify the explosion. This would be a gas explosion. For the fire setup, I used a pyro solver instead of axiom, similar to the Dunkirk effect.

After setting up the fire, it was time for the explosion! I used the pyro burst source with multiple emitters to create a large explosion. I also added pyro trail paths to refine the shape of my emitter. This process was easier than creating the procedural building HDA.

Shockwave: Again, I used a pyro burst source and employed a ray to stick it to the ground.

Blast: This was a pyro solver advected by the velocity of a rapidly expanding mask along the ground and buildings.

Debris: I customized a pyro burst source to create points moving at high velocity, then used copy-to-point with fractured box pieces to represent the debris.

Window Building Destruction: This was a challenging task that required optimization and high memory usage. I took three buildings, destroyed them, and separated the windows and other elements as colliders. I used a metaball scatter on the explosion source and re-fractured the materials. The velocity field from the blast did not advect the RBD as well as expected.

I also used topnet to create a wedging pipeline.

This project was set up in Solaris, rendered with Karma CPU, and composited in Nuke.

Showcase still 1
Showcase still 2
Showcase still 3
Showcase still 4
Showcase still 5

Comments

Sign in to join the conversation

Sign In