← ArchivePaper2016
Building and Animating User-Specific Volumetric Face Rigs
Alexandru Eugen Ichim, Ladislav Kavan, Merlin Nimier-David, Mark Pauly
System for constructing personalized volumetric facial rigs from scans, combining blendshapes with physics simulation for expressive animation.
Abstract
This paper presents a facial animation system that builds user-specific volumetric face rigs combining the direct control of blendshapes with the physical realism of simulation. From 3D scans of an actor in a neutral pose and around ten facial expressions, an average-human volumetric head template (skull, jaw, flesh tet-mesh, and skin) is registered and deformed using a Projective Dynamics optimization framework with priors such as flesh thickness measurements and a PCA face model. A volumetric extension of Example-Based Facial Rigging produces volumetric blendshapes that are driven by the same weights as traditional blendshapes but blend deformation gradients across all tetrahedra. The resulting rig delivers physics-based effects including inertia and secondary motion, volume preservation, and collision response between lips, teeth, and external objects, and the simulated model is also used to bake quadratic corrective blendshapes for fast game-engine use.
How to read this
- Category
- Method: user-specific volumetric face rigs combining blendshapes and simulation
- Contributions
- Builds personalized volumetric face rigs (skull, jaw, flesh tet-mesh, skin) from a neutral scan plus ~ten expression scans, registered with a Projective Dynamics framework and anatomical/PCA priors
- Introduces a volumetric extension of example-based facial rigging that blends deformation gradients across tetrahedra under the same weights as traditional blendshapes
- Delivers physics effects (inertia, secondary motion, volume preservation, lip/teeth/object collisions) and bakes quadratic corrective blendshapes for fast game-engine use
- Context
- Builds on the authors' avatar-from-video work (ichim-dynamic-avatar-2015), extending example-based rigging into a volumetric, physically simulated representation.Builds on: Dynamic 3D Avatar Creation from Hand-Held Video Input
- Correctness
- Assumes an average-human volumetric template can be registered to an actor from a neutral plus ~ten expressions with flesh-thickness and PCA priors; the physical realism depends on those priors and template fit, and full simulation is heavier than the baked corrective approximation it provides for runtime.
- Clarity
- Technical and pipeline-heavy; a first pass conveys the volumetric-rig concept, a second pass is needed for the Projective Dynamics fitting and the volumetric blendshape construction.
- How to read it
- First pass for the rig construction inputs and the simulation-versus-baked-correctives trade; second pass on the Projective Dynamics registration and deformation-gradient blending for implementation detail.
Builds on
Built upon by
Nothing yet.
Related work
- SoftDECA: Computationally Efficient Physics-Based Facial Animations 2023 / MIG
- Computational Bodybuilding: Anatomically-Based Modeling of Human Bodies 2015 / SIGGRAPH
- Art-Directed Muscle Simulation for High-End Facial Animation 2016 / SCA
- Realtime Performance-Driven Physical Simulation for Facial Animation 2019 / CGF
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →