A Generative AI Framework for Enhancing Human-AI Collaborative Creation in Artistic Design Services

Abstract

Amid information saturation and aesthetic pluralism, artistic design services grapple with inefficient manual workflows and imbalanced creative diversity-semantic fidelity. To address these and advance information system integration in design, this study proposes a two-stage multi-task generative AI framework for artistic design, integrating latent space remapping, hierarchical cross-modal attention distillation, and dynamic resource scheduling. Evaluated on a 30,000-sample dataset, the framework outperforms baselines: 45% lower FID than GAN-based models, 15% higher CLIP-Score for text-image alignment, over 4.3/5 professional designer satisfaction, and 1.2 iterations/ second inference on a single 3080Ti GPU. It resolves existing generative AI flaws and advances human-AI collaboration in design services, laying technical groundwork for workflow innovation, design education support, and brand development.