Generative artificial intelligence technology is profoundly transforming the paradigm of contemporary art creation. This study systematically explores the complete workflow of generative AI in art creation from conceptual conception to visual realization. The research adopts diffusion models and generative adversarial networks as core technical architectures, designing multi-dimensional art creation experiments covering key aspects such as style transfer, concept materialization, and multimodal fusion. Experimental results demonstrate that the creation system based on Stable Diffusion achieves an artistic quality score of 4.231 points (out of 5), concept-visual consistency of 87.6%, and creation time reduced by 76.3% compared to traditional methods. The study constructs an artistic creation intention encoding model, proposes a style control algorithm based on attention mechanisms, and realizes fine-grained regulation of the creation process. Through comparative experiments, we verify the differences among various generative models in artistic expressiveness, creative freedom, and technical controllability, providing theoretical support and practical guidance for the application of generative AI in the art domain. This research reveals that AI-assisted creation can significantly expand the boundaries of creative possibilities while maintaining the artist’s subjectivity, laying the foundation for the development of human-AI collaborative creation modes.
Paper
Full text
Generative AI and Contemporary Art Creation: From Concept to Visual Realization
Semantic Scholar · 2025
Abstract
Generative artificial intelligence technology is profoundly transforming the paradigm of contemporary art creation. This study systematically explores the complete workflow of generative AI in art creation from conceptual conception to visual realization. The research adopts diffusion models and generative adversarial networks as core technical architectures, designing multi-dimensional art creation experiments covering key aspects such as style transfer, concept materialization, and multimodal fusion. Experimental results demonstrate that the creation system based on Stable Diffusion achieves an artistic quality score of 4.231 points (out of 5), concept-visual consistency of 87.6%, and creation time reduced by 76.3% compared to traditional methods. The study constructs an artistic creation intention encoding model, proposes a style control algorithm based on attention mechanisms, and realizes fine-grained regulation of the creation process. Through comparative experiments, we verify the differences among various generative models in artistic expressiveness, creative freedom, and technical controllability, providing theoretical support and practical guidance for the application of generative AI in the art domain. This research reveals that AI-assisted creation can significantly expand the boundaries of creative possibilities while maintaining the artist’s subjectivity, laying the foundation for the development of human-AI collaborative creation modes.