Deep Compositional Phase Diffusion for Long Motion Sequence Generation

2025-10-16

AI Summary

Existing motion generation models struggle to maintain kinematic continuity at transition boundaries when concatenating multi-semantic motion segments, resulting in jerky artifacts. To address this, we propose a Semantic Phase Diffusion framework comprising a Semantic Phase Diffusion module and a Transition Phase Diffusion module. Operating within the latent frequency domain of a pre-trained motion phase autoencoder, our approach jointly models semantic consistency and transition-phase continuity. The framework enables end-to-end long-sequence generation, controllable motion interpolation, and intermediate-frame synthesis under fixed phase conditions. Experiments demonstrate significant improvements in semantic alignment and transition naturalness for composite motion sequences. Comprehensive evaluations across multiple benchmark tasks—including motion segmentation, interpolation, and conditional generation—validate the framework’s generality and practical effectiveness.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Recent research on motion generation has shown significant progress in generating semantically aligned motion with singular semantics. However, when employing these models to create composite sequences containing multiple semantically generated motion clips, they often struggle to preserve the continuity of motion dynamics at the transition boundaries between clips, resulting in awkward transitions and abrupt artifacts. To address these challenges, we present Compositional Phase Diffusion, which leverages the Semantic Phase Diffusion Module (SPDM) and Transitional Phase Diffusion Module (TPDM) to progressively incorporate semantic guidance and phase details from adjacent motion clips into the diffusion process. Specifically, SPDM and TPDM operate within the latent motion frequency domain established by the pre-trained Action-Centric Motion Phase Autoencoder (ACT-PAE). This allows them to learn semantically important and transition-aware phase information from variable-length motion clips during training. Experimental results demonstrate the competitive performance of our proposed framework in generating compositional motion sequences that align semantically with the input conditions, while preserving phase transitional continuity between preceding and succeeding motion clips. Additionally, motion inbetweening task is made possible by keeping the phase parameter of the input motion sequences fixed throughout the diffusion process, showcasing the potential for extending the proposed framework to accommodate various application scenarios. Codes are available at https://github.com/asdryau/TransPhase.

Problem

• Generating continuous transitions between multiple motion clips
• Preserving motion dynamics at transition boundaries between clips
• Eliminating awkward transitions and abrupt artifacts in sequences

Innovation

• Leverages dual diffusion modules for semantic and transitional guidance
• Operates in latent motion frequency domain for phase learning
• Maintains phase continuity between adjacent motion clips