HiFlow: Training-free High-Resolution Image Generation with Flow-Aligned Guidance

2025-04-08

AI Summary

High-resolution text-to-image (T2I) generation with pretrained flow-based diffusion models suffers from missing high-frequency details and degraded visual quality, while existing approaches typically require additional fine-tuning. This paper proposes a training-free framework that enhances the high-resolution generation capability of *any* pretrained flow model—without modifying model parameters or retraining. Our core innovation is the virtual reference flow mechanism, which jointly enables low-frequency-consistent initialization, structure-preserving directional alignment, and detail-faithful accelerated alignment. Grounded in flow matching principles, the method is broadly compatible with diverse T2I flow architectures—including personalized variants. Extensive experiments across multiple benchmarks demonstrate significant improvements over state-of-the-art methods, effectively unlocking the high-resolution synthesis potential of low-resolution pretrained models.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Text-to-image (T2I) diffusion/flow models have drawn considerable attention recently due to their remarkable ability to deliver flexible visual creations. Still, high-resolution image synthesis presents formidable challenges due to the scarcity and complexity of high-resolution content. To this end, we present HiFlow, a training-free and model-agnostic framework to unlock the resolution potential of pre-trained flow models. Specifically, HiFlow establishes a virtual reference flow within the high-resolution space that effectively captures the characteristics of low-resolution flow information, offering guidance for high-resolution generation through three key aspects: initialization alignment for low-frequency consistency, direction alignment for structure preservation, and acceleration alignment for detail fidelity. By leveraging this flow-aligned guidance, HiFlow substantially elevates the quality of high-resolution image synthesis of T2I models and demonstrates versatility across their personalized variants. Extensive experiments validate HiFlow's superiority in achieving superior high-resolution image quality over current state-of-the-art methods.

Problem

• Enhancing high-resolution image synthesis without training
• Aligning flow guidance for structure and detail fidelity
• Improving T2I model performance for high-resolution generation

Innovation

• Training-free high-resolution image generation framework
• Flow-aligned guidance for structure and detail
• Model-agnostic approach leveraging pre-trained flows