EXPLORING THE POTENTIAL OF GENERATIVE ARTIFICIAL INTELLIGENCE IN REVOLUTIONIZING CREATIVE WRITING AND ARTISTIC EXPRESSION

Abstract

Gen-AI technologies like LLMs and multimodal generative networks have transformed creative writing and art. Gen-AI systems like GPT-4, Claude 3, Stable Diffusion XL, and Midjourney v6 may create syntactically coherent, semantically rich, and visually engaging creative creations. Traditional computational creativity tools used rule-based heuristics and shallow statistical models Despite these advances, generated content sometimes lacks narrative coherence, stylistic inconsistencies, originality, and IP attribution. To circumvent these limitations, Creative Contextual Generative Architecture (CCGA) uses a transformer-based semantic controller, style-conditioned variational encoder, and human-in-the-loop (HITL) feedback module. The suggested system has creative generation and quality arbitration pipelines. In narrative fiction, lyrical poetry, and visual art synthesis, CCGA has a BLEU-4 score of 0.748, a ROUGE-L F1 of 0.812, a Frechet Inception Distance (FID) of 9.43, and an 87.4% human preference rate above baseline systems. The framework dominates GPT-4-only, DALL-E 3-only, and fine-tuned BERT baselines. A reproducible evaluation process, a new benchmark dataset (CreativeAI-Bench 2025), and practical design concepts for responsible Gen-AI implementation in creative sectors are presented in this paper