UniCSG: Unified High-Fidelity Content-Constrained Style-Driven Generation via Staged Semantic and Frequency Disentanglement

TL;DR

UniCSG introduces a staged training framework for high-fidelity style transfer that effectively disentangles content and style, improving robustness and content preservation in diffusion-based models.

Contribution

It proposes a novel staged training approach with frequency-aware disentanglement and pixel-space reward learning for better style transfer performance.

Findings

Enhanced content faithfulness and style alignment demonstrated in experiments.

Improved robustness and stability in style-driven generation.

Effective disentanglement of content and style in diffusion models.

Abstract

Style transfer must match a target style while preserving content semantics. DiT-based diffusion models often suffer from content-style entanglement, leading to reference-content leakage and unstable generation. We present UniCSG, a unified framework for content-constrained, style-driven generation in both text-guided and reference-guided settings. UniCSG employs staged training: (i) a latent-space semantic disentanglement stage that combines low-frequency preprocessing with conditioning corruption to encourage content-style separation, and (ii) a latent-space frequency-aware detail reconstruction stage that refines details via multi-scale frequency supervision. We further incorporate pixel-space reward learning to align latent objectives with perceptual quality after decoding. Experiments demonstrate improved content faithfulness, style alignment, and robustness in both settings.