HairCS: Reconstructing Strand-Based Hair from Hair Cards

HairCS is an automated pipeline for converting hair-card models into strand-based hairstyles. The goal is to preserve the authored look of the original cards while generating denser, more detailed strand geometry that can feed modern hair rendering and simulation setups.

The system is designed around production constraints rather than a purely academic reconstruction. Strands are generated to originate from the scalp, roots are distributed uniformly, and the resulting volume is filled plausibly, which matters if the asset needs to survive grooming edits instead of just looking good in a demo.

That makes the work especially relevant for character teams that still rely on cards for authoring efficiency but want strand-based output for higher-end rendering, physics, and grooming modifiers such as clumping, curling, and noise. In practice, that could reduce the manual labor of rebuilding hair for next-gen pipelines while keeping artistic control intact.

The method was validated across a broad range of hairstyles, including short cuts, long hair, curly styles, buns, and ponytails. The authors are from the University of Utah, LIGHTSPEED, and UCLA, and the work was submitted on September 15, 2026.