Tearing Apart

Responsibilities: 3D Modeling, Vellum tearing simulation, VEX coding, Vector math, Rendering

Overview

Tearing Apart is a project that combines dynamic vellum simulation with creative coding to bring cloth to life in an engaging way. By utilizing weld and cloth constraints, additional forces, and VEX coding, the project simulates the physical behavior of torn cloth. The process begins with creating the base geometry, followed by adding noise to the positions and edges of the cloth, setting up constraints to simulate tearing, and finally, enhancing the visual impact through code and custom materials. This results in an immersive experience where cloth pieces break apart and swirl in the air, highlighting the chaotic beauty of their motion.

3D modeling

The vellum pieces start with a simple 60x30 grid as the base geometry. I remeshed the grid geometry to create a solid foundation for further transformations using a for-loop to give the cloth a more natural look, especially around the torn edges and disconnected pieces. Each cloth element's position was altered with noise, and I distorted the edges and randomly rotated each piece, adding randomness to the design and enhancing its organic appearance.

VEX coding

To add a more cartoonish and visually engaging effect to the vellum cloth, I used VEX coding to create the "outlines." The first step was calculating the val attribute by taking the dot product between the normal vector at each point and the direction vector dir that points from the point to the camera. The direction vector is derived by subtracting the camera's position from each point's position, then normalizing it. The result of this dot product is a value between -1 and 1, where 1 indicates the normal is facing directly at the camera, 0 means it's perpendicular to the camera's view, and -1 indicates the normal is facing away. I focused on values close to zero, where the normal is nearly perpendicular to the camera, which are typically found along the edges or folds of the cloth.

In the second part of the code, I used a threshold value to isolate the points where the normal is nearly perpendicular to the camera. Points where the val attribute was too far from zero were filtered out, either by removing values below the negative threshold or above the positive threshold. This left only the points where the normal was almost perpendicular to the camera, typically outlining concavities in the cloth.

Turbulent Behavior Simulation with Vellum Solver

In the Vellum solver, I reduced gravity to make the cloth pieces appear to levitate. To enhance their movement, I introduced a Pop Axis force, which applied a vertical force to make the pieces circulate around a defined shape and rise. I then added wind to introduce an unpredictable, swirling behavior that gave the cloth a more chaotic, turbulent motion.

Tearing of the Cloth Pieces Through Vellum Constraints

To make the cloth tear naturally as it swirled, I used Vellum weld constraints to stitch the pieces, produced by the Edge Fracture node, together. I set a breaking threshold of 0.0004 to ensure the pieces would tear apart when forces started to stretch them beyond their limits. As the geometry moved and forces acted on it, the cloth naturally began to tear, adding realism to the simulation.

Rendering

For rendering, I created three custom materials: one for the cloth, one for the lines on the cloth, and one for the background. I designed the background to contain shadows from the flying cloth pieces, so I created a highly diffusive orange material. To ensure no reflections from the lighting, I set the roughness of the reflection to 0, keeping it matte. For the black lines outlining the cloth, I rendered them as strands and adjusted their scale and type to achieve the desired visual effect.

Challenges and learning outcomes

Challenges and Learning Outcomes One of the challenges I faced was creating unique ripped-edge effects for each piece of cloth.

Initially, since all the cloth pieces were identical, applying noise and fracture produced the same result for every piece. To solve this, I utilized the "iteration" attribute generated by the for-loop, which is unique for each piece. By setting the offset for noise through the equation rand(detail(0, "iteration", 0)), I generated a pseudorandom value based on the iteration attribute. Similarly, for the Edge Fracture node, I defined the number of pieces the cloth fractured into based on detail(0, "iteration", 0)/5, which created a value from 0 to 5 depending on the iteration. This approach allowed each piece to have unique transformations, making the final result more varied and realistic.

Below are the effects of these transformations on different pieces of cloth.