AI image generation technology for visual communication design

This study addresses two key challenges in applying artificial intelligence (AI) image generation to visual communication design: insufficient alignment with design requirements and unstable style controllability. It systematically explores the integration of AI technology with professional design practice. First, the current research and application status of AI generation in design is reviewed, and its technological positioning is clarified. Next, an integrated framework is constructed by combining diffusion models, CLIP-based cross-modal alignment, and ControlNet spatial constraints. This framework supports an AI-driven design workflow that incorporates requirement analysis, controllable generation, and multi-solution exploration. Multi-dimensional evaluation metrics, including generation quality, peak signal-to-noise ratio, and requirement matching, are also established. Experiments are conducted across the core design scenarios-posters, brand logos, and mobile interfaces-using public datasets such as DesignBench-20 K and a custom commercial test set. The model performs best in mobile interface scenarios. Across all scenarios, the design cycle is reduced by over 91%. The model reliably generates outputs under textual requirement noise ≤ 10% and non-design data distribution shifts ≤ 40%. Ablation studies highlight the critical and differentiated contributions of each core component in layout control and requirement matching. The study further clarifies the application boundaries of this technology in visual communication design. It proposes optimization paths, including enhancing CLIP's abstract semantic feature extraction, building a ControlNet template library for design-specific constraints, and refining GAN style interpolation weight adjustment. Finally, the study identifies limitations in data adaptability and dynamic generation capability, providing directions for future improvement.