High-Fidelity Mobile Avatars with Pruned Local Blendshapes

2026-05-03

AI Summary

This work addresses the challenge of real-time, high-fidelity rendering of human avatars on mobile devices by proposing a Gaussian body representation based on localized linear blendshapes. The method models pose-dependent, nonlinear variations in Gaussian attributes within local regions and prunes blendshapes with negligible deformation contributions to achieve significant model compression while preserving fine geometric details. The system enables end-to-end training without requiring any pre-trained models and leverages WebGPU for cross-platform deployment. Experimental results demonstrate that the proposed approach achieves real-time rendering at 120 FPS under 2K resolution on mobile hardware, outperforming existing mobile-oriented methods in terms of visual fidelity.

Abstract

We propose a method to reconstruct high-fidelity human avatars from multi-view video that can run on mobile devices. Many works can model high-quality Gaussian-based full-body avatars from multi-view video. However, these methods require heavy computation to obtain pose-dependent appearance, making deployment on mobile devices very difficult. Recent methods distill from pretrained models and model pose-dependent nonlinear Gaussian attributes by linearly combining global pose features with blendshapes. Although they can run on mobile devices, they suffer some loss of detail. We observe that nearby Gaussians are often highly correlated within a local region of the body, and can be linearly modeled with less error. Therefore, we use local linear blendshapes in small body parts to capture global nonlinear changes of Gaussian attributes. To further reduce computation and model size, we propose to remove blendshapes for Gaussians whose attributes change little, yielding a minimal blendshape representation. Our method is an end-to-end training method without a pretrained model. To make it run on multiple devices, we implement our method using WebGPU. Experiments show that our method can render high-quality human avatars with better details, and can reach 120 FPS at 2K resolution on mobile devices.

Problem

• high-fidelity avatars
• mobile deployment
• pose-dependent appearance
• Gaussian-based rendering
• blendshapes

Innovation

• local blendshapes
• pruning
• Gaussian avatars
• mobile rendering
• WebGPU