With the rise of deep learning and generative confrontation networks, the application of machine vision algorithms to multimedia digital AI painting systems has great potential and room for innovation. This article designs a multimedia digital artificial intelligence painting system. This system focuses on machine vision algorithms and uses generative adverse network algorithm and model implementation to realize automatic painting generation and image repair and enhancement. The construction of the system mainly includes key steps such as data collection, GAN model selection and training, adjustment of hyperparameters and selection of evaluation metrics. After testing, it is found that the training time of the system in this paper is between 6–14 s. 170–176 the system can automatically generate beautiful works of art, repair damaged images and improve the quality and perception effect of images.