Plan, Don't Pose: Long Composite Motion Generation with Text-Aligned BFM

2026-05-28

AI Summary

This work addresses the limitations of existing text-to-motion generation approaches, which struggle with long-horizon, semantically complex instructions due to the tight coupling of semantic understanding, structural planning, and motion synthesis within a single model. To overcome this, we propose Text2BFM, a novel framework that aligns natural language with a frozen pre-trained Behavior Foundation Model (BFM) for the first time. By introducing a text-guided variational behavior bottleneck, our method compresses policy sequences within the BFM’s latent policy space, effectively decoupling high-level semantic planning from low-level motion execution. Notably, Text2BFM avoids end-to-end retraining of the motion generator; instead, it employs a lightweight conditional generator to synthesize motions on a compact behavior manifold. This design significantly enhances robustness, efficiency, and motion quality for long, composite actions, outperforming current end-to-end methods.

Abstract

Text-to-motion (T2M) generation has broad applications in character animation, virtual avatars, and human-robot interaction. Existing methods typically generate pose trajectories or motion tokens directly from language, forcing a single model to handle semantic interpretation, long-horizon structure, and low-level physical realization. This coupling makes them costly and often unreliable for long, compositional, or semantically dense prompts. We propose Text2BFM, the first framework that aligns natural language with pretrained Behavioral Foundation Models (BFMs) for T2M generation without relying on heavy end-to-end motion generators. Text2BFM operates in the latent policy space of a frozen BFM, using it as an executable motion prior. A text-aligned variational behavioral bottleneck compresses BFM policy-latent sequences into compact motion representations that are compatible with language and preserve long-horizon behavioral structure. Generation is performed in this compact behavioral manifold with a lightweight conditional generator, and the resulting latent encoded behaviors are decoded into policy latents that drive the pretrained frozen BFM. By decoupling semantic planning from motion execution, Text2BFM achieves efficient, robust T2M generation and strong performance on long, compositional textual descriptions.

Problem

• text-to-motion
• long-horizon motion
• compositional prompts
• semantic alignment
• motion generation

Innovation

• Behavioral Foundation Models
• text-to-motion generation
• semantic-motion decoupling
• variational behavioral bottleneck
• latent policy space