As a time-honored and technically complex art form, oil painting is undergoing unprecedented transformation in its creative process and aesthetic value under the influence of AI technology. This study explores the modernization and innovation of oil painting creation from an AI technology perspective, as well as its role in driving artistic development. It primarily achieves oil painting style transfer through the optimization of the CycleGAN (Cycle-Consistent Generative Adversarial Network), proposes the introduction of spectral normalization processing, and improves the residual structure. Finally, the algorithmic model is validated through quantitative and qualitative experiments. The experiments demonstrate that the model generates oil painting images with outstanding performance in terms of clarity, diversity, and style similarity. Its IS values are 4.17% to 172.02% higher than those of the comparison methods, while its ID values are reduced by 2.99% to 63.50%. Additionally, the subjective quality evaluation and style similarity evaluation of the generated images are optimal. The model constructed in this paper improves the visual effects of image style transfer and can be used to assist artists in oil painting creation, inspiring their creative inspiration.