UNDER CONSTRUCTION - Transpacific Express Main Character

Transpacific Express is a Montréal-based collective operated by El Jousse, Frantz Lin, and Celia Benhocine. Rooted in Asian and African solidarity movements of the 1960s, TE curates and performs work exploring connections between BIPOC youth cultures, collaborating with organizations such as MAI, YATAI MTL, and RIDM.

Wanting to explore a more interdisciplnary media practice, TE dreamt up their digital avatar. They wanted a "main character", essentially a fourth member of the collective, to create a wide range of content with the aim of engaging the community; video essays, TiKTok dances, hosting digital events, etc.!

I was initially brought on as a rigger, and later stepped into the role of Technical Director. My early work focused on building the character rig, along with outfits and props. As the project evolved, so did my responsibilities! TE needed something accessible, versatile, and easy for non-techy people to pick up and run with. Using VTuber tech and worklows was a natural conclusion!

Though adopting VTuber seemed like a stroke of genius, the process was far from straightforward. The available tools weren’t designed for the level of visual fidelity we were aiming for, and many standard game pipelines simply didn’t translate. A lot of the work involved trial and error; debugging broken exports, rebuilding rigs to survive format constraints, and finding workarounds for systems that would silently fail or strip critical data. At the same time, TE wanted to avoid an overly stylized, cartoon-y look. This created a constant tension between visual goals and technical limitations, especially within rendering systems that don’t natively support PBR workflows!

The core challenge became: how do we achieve a PBR-like aesthetic within systems that don’t natively support it? To solve this, I rebuilt the pipeline from the ground up around those constraints. I established modeling and texturing guidelines and budgets to prevent downstream breakage, and ensured assets were authored with export limitations in mind from the start. I handled all rigging and technical animation, designing setups that could survive repeated conversion without degrading. I implemented a full set of ARKit blendshapes to drive facial performance, carefully balancing fidelity with runtime constraints. On the rendering side, I developed custom HLSL shaders to approximate PBR behavior, compensating for missing lighting data and limited shading models.

The result is a continuously evolving pipeline that delivers a grounded, semi-realistic look despite the limitations of the original toolset, with ongoing work to expand features and push those constraints further. This has been such a fun learning experience, and VTuber tech is still in its infancy! I am looking forward to developing this concept further, and continuing to push the limits!