FunPhase: A Periodic Functional Autoencoder for Motion Generation via Phase Manifolds

Learning natural body motion remains challenging due to the strong coupling between spatial geometry and temporal dynamics. Embedding motion in phase manifolds, latent spaces that capture local periodicity, has proven effective for motion prediction; however, existing approaches are tied to fixed skeletons and narrow motion distributions, limiting their applicability across diverse settings. We introduce FunPhase, a functional periodic autoencoder that learns a phase manifold for motion and replaces discrete temporal decoding with a function-space formulation, enabling smooth trajectories that can be sampled at arbitrary temporal resolutions. FunPhase unifies motion prediction and generation within a single interpretable phase manifold, enabling motion generation via latent diffusion, generalizes across skeletons and datasets, and supports downstream tasks such as motion super-resolution and partial-body completion. Our model achieves substantially lower reconstruction error than prior periodic autoencoder baselines, achieving uniform improvements of at least 45% across all metrics, while enabling a broader range of applications and performing on par with state-of-the-art motion generation methods.

Creating realistic animated movement for characters in video games, films, and virtual avatars is notoriously difficult. Existing AI methods often struggle because they treat motion as a series of disconnected snapshots rather than smooth, flowing action, often producing glitches like feet sliding on the floor or jittery transitions. Moreover, the most successful prior approaches only work for one specific skeleton (like a human), so an animator working with a dog, bird, or fantasy creature has to start from scratch. We developed FunPhase, a system that learns to represent motion as a continuous mathematical function rather than a fixed sequence of frames. The motion representation is constrained to be periodic, naturally capturing the rhythms hidden in movement (like the cyclical pattern of a walking gait) which acts as a built-in guide toward physically believable results. Crucially, FunPhase is skeleton-agnostic, meaning the same model can learn from humans, dogs, and dozens of other creatures. FunPhase cuts reconstruction errors by at least 45% compared to previous phase-based methods and matches state-of-the-art generation quality. Because motion is stored as a continuous function, animators can fill in missing body parts, smooth sparse keyframes into full sequences, or replay motion at any frame rate, opening up flexible new tools for animation, games, and simulation.