ASTRA: Enhancing Multi-Subject Generation with Retrieval-Augmented Pose Guidance and Disentangled Position Embedding

TL;DR

ASTRA is a novel framework that improves multi-subject image generation by disentangling appearance and pose using retrieval augmentation and specialized positional encoding within a diffusion transformer.

Contribution

It introduces a dual-strategy approach combining retrieval-augmented pose priors and a new asymmetric position embedding to better preserve identity and pose in generated images.

Findings

Achieves state-of-the-art pose adherence on COCO-based benchmark.

Maintains high identity fidelity and text alignment in experiments.

Outperforms existing methods in complex multi-subject pose generation.

Abstract

Subject-driven image generation has shown great success in creating personalized content, but its capabilities are largely confined to single subjects in common poses. Current approaches face a fundamental conflict when handling multiple subjects with complex, distinct actions: preserving individual identities while enforcing precise pose structures. This challenge often leads to identity fusion and pose distortion, as appearance and structure signals become entangled within the model's architecture. To resolve this conflict, we introduce ASTRA(Adaptive Synthesis through Targeted Retrieval Augmentation), a novel framework that architecturally disentangles subject appearance from pose structure within a unified Diffusion Transformer. ASTRA achieves this through a dual-pronged strategy. It first employs a Retrieval-Augmented Pose (RAG-Pose) pipeline to provide a clean, explicit structural prior from a curated database. Then, its core generative model learns to process these dual visual conditions using our Enhanced Universal Rotary Position Embedding (EURoPE), an asymmetric encoding mechanism that decouples identity tokens from spatial locations while binding pose tokens to the canvas. Concurrently, a Disentangled Semantic Modulation (DSM) adapter offloads the identity preservation task into the text conditioning stream. Extensive experiments demonstrate that our integrated approach achieves superior disentanglement. On our designed COCO-based complex pose benchmark, ASTRA achieves a new state-of-the-art in pose adherence, while maintaining high identity fidelity and text alignment in DreamBench.