PRISM: Streaming Human Motion Generation with Per-Joint Latent Decomposition

TL;DR

PRISM introduces a structured joint-factorized latent space and noise-free condition injection, enabling high-quality, versatile, and streaming human motion generation from text and pose inputs, with improved long-horizon synthesis.

Contribution

It proposes a novel joint-factorized latent space and a noise-free conditioning method, unifying multiple motion generation tasks in a single model.

Findings

Achieves state-of-the-art results on multiple benchmarks.

Enables seamless text-to-motion and pose-conditioned generation.

Supports autoregressive streaming synthesis with reduced drift.

Abstract

Text-to-motion generation has advanced rapidly, yet two challenges persist. First, existing motion autoencoders compress each frame into a single monolithic latent vector, entangling trajectory and per-joint rotations in an unstructured representation that downstream generators struggle to model faithfully. Second, text-to-motion, pose-conditioned generation, and long-horizon sequential synthesis typically require separate models or task-specific mechanisms, with autoregressive approaches suffering from severe error accumulation over extended rollouts. We present PRISM, addressing each challenge with a dedicated contribution. (1) A joint-factorized motion latent space: each body joint occupies its own token, forming a structured 2D grid (time joints) compressed by a causal VAE with forward-kinematics supervision. This simple change to the latent space -- without modifying the generator -- substantially improves generation quality, revealing that latent space design has been an underestimated bottleneck. (2) Noise-free condition injection: each latent token carries its own timestep embedding, allowing conditioning frames to be injected as clean tokens (timestep0) while the remaining tokens are denoised. This unifies text-to-motion and pose-conditioned generation in a single model, and directly enables autoregressive segment chaining for streaming synthesis. Self-forcing training further suppresses drift in long rollouts. With these two components, we train a single motion generation foundation model that seamlessly handles text-to-motion, pose-conditioned generation, autoregressive sequential generation, and narrative motion composition, achieving state-of-the-art on HumanML3D, MotionHub, BABEL, and a 50-scenario user study.