GALA: Gaussian Blendshape Distillation for Real-Time Avatars
What it is
This paper introduces GALA, a distillation method that replaces the costly neural inference in 3D Gaussian avatars with a linear combination of identity-independent blendshapes. It uses a shallow MLP to predict blendshape coefficients and a linear blend to animate avatars, significantly reducing computational overhead while maintaining rendering quality.
The problem it solves
Real-time animation of 3D Gaussian avatars is bottlenecked by the high computational cost of per-frame neural inference, preventing efficient deployment on resource-constrained devices like mobile phones.
What you could use it for
- Real-time facial expression animation for 3D Gaussian avatars
- Full-body animation with clothing dynamics on mobile devices
- Accelerating inference for existing pretrained avatar models without retraining
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.