DynHair at ECCV 2026: Dynamic 3D Hair Strand Reconstruction for Photorealistic Head Avatars

Michael_J_Black · x · 2026-09-09

Researchers from ETH Zurich, MPI and others present DynHair (ECCV 2026): reconstructing dynamic explicit hair strands from multi-view video using structured 3D Gaussian Splatting. A temporal network conditioned on head angular velocity, acceleration and gravity models strand deformation, jointly optimized with the face-region 3DGS via differentiable splatting. Paper and code are available.

Original post →

More from Multimodal

Multimodal channel →