# Apple — GHARP: Real-Time Gaussian Head Animation from Large-Scale Reconstruction Prior

- Company: Apple (apple.com)
- Announced: 2026-10-26
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://machinelearning.apple.com/research/gharp-head-animation
- Record: https://forck.live/items/27961-gharp-real-time-gaussian-head-animation-from-large-scale-reconstruction-prior
- Subject: Machine Learning Research

Apple researchers present GHARP, a method that animates 3D human heads in real time from a few input images of a subject and a driving expression signal. The work splits the problem into an identity stage that builds the subject's geometry and appearance offline and an animation stage that predicts expression-dependent residuals at runtime. That split lets the identity stage be expensive while the animation stage runs a lightweight network optimized for mobile devices, which the authors say gives a favorable trade-off across fidelity, quality and runtime. Animation happens in a semantically structured latent space of a pretrained reconstruction model, where expression changes stay spatially contained and residual prediction stays efficient. The reconstruction prior also gives a consistent spatial layout, letting the method fuse multiple input views into a compact, fixed-size canonical Gaussian representation. The authors note that expression codes describe only the face, so body pose and clothing position are underspecified and the animation network averages over conflicting body appearances, producing blur and temporal flicker. A body alignment network aligns the person's body in the target image with the input reference images to remove that ambiguity from the training signal. The authors report state-of-the-art quality on the Ava-256 benchmark while running up to 13× faster on an A100 GPU with 8× fewer Gaussians.

## Evidence

Verbatim from https://machinelearning.apple.com/research/gharp-head-animation:

> Authors Ali 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.

---

Record: https://forck.live/items/27961-gharp-real-time-gaussian-head-animation-from-large-scale-reconstruction-prior
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md