Artistic Simulation of Curly Hair

How to read this

Category

Production hair simulation method (artist-directed curly hair)

Contributions

• Introduces a hair model built on a mass-spring system with a novel bending spring that computes stable material frames along a smoothed version of the hair curve, avoiding the unwanted rotation that elastic-rod models introduce during simple motion like a walk cycle.
• Adds a separate core spring that controls longitudinal stretch of the curl independently from bending stiffness, so curls hold their helical shape under extreme acceleration without needing to stiffen the bend and lose artistic flexibility.
• Presents a hair-hair contact pruning algorithm that reduces the number of contact pairs and particles considered, making dense curly hair simulation parallelizable and fast enough for production schedules.
• Reports production use on two full-length Pixar features across a range of hair styles from straight to curly, with results usable without extensive per-shot parameter tuning.

Context

The paper explicitly departs from physically accurate elastic-rod curl models like Bergou et al.'s 2008 Discrete Elastic Rods, which the authors show introduces visible unwanted rotation in a curl during ordinary motion, in favor of a purpose-built mass-spring model that prioritizes the artistic look over strict physical correctness. It sits in the mass-spring hair simulation lineage going back to Rosenblum et al. 1991 and Petrovic et al. 2005, but is notable for explicitly choosing non-physical behavior when physical accuracy conflicts with what animators want to see.Builds on: Discrete Elastic Rods

Correctness

The claims rest on production validation across two full Pixar features rather than a physical ground-truth comparison, which is appropriate given the paper's explicit goal of artistic control over physical fidelity, the authors are candid that their bending model deliberately avoids the physically correct rotational behavior of elastic rod models. Its main limitation is that it targets guide hair simulation for stylized looks, not the dense strand-level or fully physical hair simulation used in other production or research contexts, and the parallel contact pruning is a performance technique that trades some accuracy for speed.

Clarity

A well written, practically oriented SCA paper, the figures comparing elastic-rod rotation artifacts against this method's stable bending are especially clear. It is accessible to any CFX or hair TD without requiring heavy physics background beyond basic spring mechanics.

How to read it

First pass, read the abstract and look at Figure 2's comparison between the elastic rods model and this paper's bending model on a simple curl under motion, that single figure is the paper's core motivation. Second pass, read sections 3.2 and 3.3 on the bending spring and core spring formulations, since those two additions are the specific technical contribution over prior mass-spring hair work. Third pass, read the contact pruning section if performance at scale is your concern, and compare its approach to whatever hair-hair collision handling your current pipeline uses, since GPU-accelerated grooming tools have since taken over much of this problem in production.