Cem Yuksel, Computer Graphics Professor at the University of Utah, has announced an impressive, if not astonishing, achievement for real-time hair rendering. The new method achieved by his team can efficiently render strand-based hair models on the GPU, resulting in a significant reduction in storage and memory bandwidth.
In the article introducing the project, a GIF at the top shows that their technique can render a scene with 100 characters, each with a unique hair model comprised of 100,000 strands, in just 2 milliseconds on an NVIDIA GTX 4090 GPU, using a mere 1.7 MB of storage for all the models.