taoavatar-augmented-reality - Real-Time Talking Avatars for AR
Create real-time full-body talking avatars for augmented reality with hybrid SMPLX-Gaussian representations and teacher-student deformation distillation.
Tags
Updated: 2026-09-30Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Build hybrid SMPLX-Gaussian models
- Predict pose-dependent deformations
- Distill teacher deformations
- Control avatar expressions and gestures
- Synchronize avatars with audio
- Render Gaussian avatars in real time
Inputs
- SMPLX body model
- Pose parameters
- Shape parameters
- Expression parameters
- Hand pose parameters
- Texture features
- UV maps
- Gesture keyframes
- Audio signals
Outputs
- Rendered full-body talking avatars
- Pose-dependent deformation predictions
- Compact student deformation model
- Gaussian rendering parameters
Requirements
- PyTorch environment
- PyTorch3D
- Pretrained SMPLX model
- 3D Gaussian rasterization support