Hosuk Chang
Personality Pavilions for Soul
This technique originated as a way to create atmospheric color gradients in 3D space, essentially animating a 2D polygon grid with shifting hues and capturing the color values in volume. The resulting effect, like a very, very long motion blur, accidentally produced shapes similar to the Personality Pavilions we were designing. This happy accident led us to adopt the technique for those structures.
Because the process required animating meshes, the pavilions were largely designed directly in 3D, rather than the traditional 2D-to-3D workflow. It was one of the most challenging, yet inspiring, projects I've worked on, because it meant not only tackling significant technical challenges, but also finding a way to represent abstract concepts – human personalities such as insecurity, humour, excitement, and even self-absorption – in visually compelling ways.
Unlike the static pavilions, some, such as this Insecure Pavilion, were animated. This involved generating distinct polygon grid animations for each individual frame.
The final pavilion models' 3D volume didn't capture color values directly. Instead, they captured UVW coordinates. The U and V values came from the polygon mesh's UVs, while the W value represented normalized time. This allowed for remapping with custom ramp parameters in the volume shader, providing greater control over color and enabling additional volume displacement textures.
The design of some pavilions, like the Funny Pavilion, demanded a more complex approach. It wasn't immediately clear how to animate the polygon grid. In this case, I animated the underlying meshes to create thickness and then captured specially designed UVs on each mesh. This allowed for edge feathering and subtle color shifts, resulting in a more sophisticated final look.
The animated pavilions, such as the Excited Pavilion shown here, were first visualized through polygon mesh animation. This allowed for rapid prototyping and easy iteration. Once the motion was approved, the design was finalized by generating the 3D volume.