HyperGaussians: High-Dimensional Gaussian Splatting for High-Fidelity Animatable Face Avatars

TL;DR: Plug-and-play high-dimensional extension of 3DGS for dynamic scenes.

We introduce HyperGaussians, a novel extension of 3D Gaussian Splatting for high-quality animatable face avatars. While tremendous successes have been achieved for static faces, animatable avatars from dynamic videos still fall in the uncanny valley. The de facto standard, 3D Gaussian Splatting (3DGS), represents a face through a collection of 3D Gaussian primitives. 3DGS excels at rendering static faces, but the state-of-the-art still struggles with nonlinear deformations, complex lighting effects, and fine details. While most related works focus on predicting better Gaussian parameters from expression codes, we rethink the 3D Gaussian representation itself and how to make it more expressive. Our insights lead to a novel extension of 3D Gaussians to high-dimensional multivariate Gaussians, dubbed 'HyperGaussians'. The higher dimensionality increases expressivity through conditioning on a learnable local embedding. However, splatting HyperGaussians is computationally expensive because it requires inverting a high-dimensional covariance matrix. We solve this by reparameterizing the covariance matrix, dubbed the 'inverse covariance trick'. This trick boosts the efficiency so that HyperGaussians can be seamlessly integrated into existing models. To demonstrate this, we plug in HyperGaussians into two state-of-the-art methods for face avatars: FlashAvatar and GaussianHeadAvatar. Our evaluation on 29 subjects from 6 face datasets shows that HyperGaussians outperform 3DGS numerically and visually, particularly for high-frequency details like eyes, teeth, wrinkles, and specular reflections.

Please use the links in the navigation bar to quickly jump to results.

We recommend using Chrome for playing the videos.

We propose an extension to 3D Gaussians, dubbed HyperGaussians, and plug them into an existing method for face avatars, FlashAvatar. FlashAvatar modulates 3D Gaussian primitives with expression-dependent offsets. We make a single modification to the pipeline: We plug in HyperGaussians between the MLP output and the rasterization, which modifies the offsets in higher dimensions. Instead of directly predicting offsets , we predict a latent that conditions the HyperGaussians. Without any other modifications or hyperparameter tuning, this simple change enhances details in the final avatar and leads to a performance boost. This figure has been adapted from FlashAvatar.

Qualitative comparison in the multi-view video setting against GaussianHeadAvatar (GHA) [Xu et al. 2024]. Move the cursor over the images to compare. Top half: GaussianHeadAvatar (left) vs. Ours (right). Bottom half: Ground Truth.

HyperGaussians achieve high-quality details for thin structures (glass frames and teeth in the first row), specular reflections (eyes in the second row), and gracefully handle complex deformations (mouth in the third row).

HyperGaussians preserve fine details in the teeth and the overall shape of the subject.

HyperGaussians demonstrate an emerging property that arises naturally throughout training. Their conditional covariance matrices indicate the variance of each Gaussian across the different expressions of the training subject and can be intuitively interpreted as uncertainty. This agreement between the uncertainty estimates and what would intuitively be considered difficult regions is an inductive bias of our formulation and does not require explicit supervision.