Cel shader (left) vs PBR adaptation (right) — Sunflower, P1 Games 2024
The Problem
Stylised vs. physically grounded
Team Mighty Hibiscus initially agreed on a hand-drawn cel aesthetic for Sunflower. After implementing the cel shader, playtesting revealed that the flat look conflicted with UE5's lighting system, causing characters to read poorly against the environment. I had to evaluate the original design decision and propose a shift to PBR mid-jam.
What I Built
What changed and why
- Original cel shader — step-based diffuse banding with outlined edges using post-process depth and normal edge detection
- The problem — flat shading broke UE5's dynamic lighting and made the scene read as two disconnected visual layers
- PBR adaptation — retained stylised colour palette and toon-like roughness values while working with, not against, the engine's lighting model
- Design decision — documented the trade-off for the team: losing the hand-drawn feel but gaining visual coherence and faster lighting iteration
Reflection
What this taught me
The most valuable part of this project wasn't the shader itself — it was learning to reverse a design decision under deadline pressure and communicate that change clearly to a team. The technical implementation was secondary to the design reasoning.
Research Question
When does stylised rendering work against an engine's pipeline rather than with it — and how should that constraint inform early aesthetic decisions in game production?