GHARP brings real-time Gaussian head animation to mobile
GHARP is a method for animating 3D human heads in real time from a few images of a subject and a driving expression signal. Its authors say it separates an offline identity-reconstruction stage from a lightweight runtime animation stage optimized for mobile devices. The runtime predicts expression-dependent residuals in the semantically structured latent space of a pretrained reconstruction model. The authors also report that ambiguous body appearance can cause blur and temporal flicker, which they address with a body alignment network. They claim state-of-the-art quality on Ava-256, execution up to 13 times faster on an A100 GPU, and eight times fewer Gaussians; the supplied evidence does not independently verify these results.
The reported two-stage design permits an expensive offline identity stage while using a lightweight animation network optimized for mobile devices, explicitly improving the runtime-quality trade-off. The performance and quality figures remain author-reported and were not independently verified in the supplied evidence.
research area Computer Visionconference Asian Conference on Computer Vision (ACCV)
content type paperpublished October 2026
GHARP: Real-Time Gaussian Head Animation from Large-Scale Reconstruction Prior
AuthorsAli Benlalah*, Sepehr Johari*, Patricia Vitoria*, Armin Kappeler*, Artem Sevastopolsky, Alexander Jung, Gabriele Fanelli, Kevin Mader, Manuel Breitenstein, Claudia Plüss, Jan Rüegg, Simon Biland, Thomas Etterlin, Dmitry Kostiaev, Mathias Deschler, Brian Amberg, Sebastian Martin
We present GHARP (Real-time Gaussian Head Animation from Large-scale Reconstruction Prior), a method that animates 3D human heads in real time from a few input images of a subject and a driving expression signal. We decouple the problem into an identity stage that builds a representation of the subject’s geometry and appearance offline, and an animation stage that predicts expression-dependent residuals on top of it at runtime. This separation offers a favorable trade-off with respect to fidelity, quality and runtime: the identity stage can be expensive while the animation stage runs a lightweight network, optimized for mobile devices. Our method performs animation in a semantically structured latent space of a pretrained reconstruction model, where expression changes remain spatially contained, making residual prediction efficient. This reconstruction prior provides a consistent spatial layout, allowing fusion of multiple input views into a compact, fixed-size canonical Gaussian representation. While this two-stage design improves the runtime-quality trade-off, it still inherits a problem common to all expression-driven avatar methods: expression codes describe only the face and thus omit body pose and clothing position, making these regions underspecified in the input. The animation network faces an ill-posed mapping and resorts to averaging over conflicting body appearances, producing blur and temporal flicker. We address this with a body alignment network that learns to align the person’s body in the target image with the input reference images, removing the ambiguity from the training signal. Our method achieves state-of-the-art quality on the Ava-256 benchmark while running up to 13× faster on an A100 GPU with 8× fewer Gaussians.
#GHARP#Ava-256#NVIDIA A100#3D Gaussian Splatting#Apple Machine Learning Research