transports the natural Bishop frame directly, similar to the reference frame used in [Bergou et al. 2008]. However, we compute this frame along a smoothed representation of the hair curve instead of the curve itself, reducing the sensitivity of the frame to changes in hair positions.