Digital arts have gained an unprecedented level of popularity with the emergence of non-fungible tokens (NFTs). NFTs are cryptographic assets that are stored on blockchain networks and represent a digital certificate of ownership that cannot be forged. NFTs can be incorporated into a smart contract which allows the owner to benefit from a future sale percentage. While digital art producers can benefit immensely with NFTs, their production is time consuming. Therefore, this paper explores the possibility of using generative adversarial networks (GANs) for automatic generation of digital arts. GANs are deep learning architectures that are widely and effectively used for synthesis of audio, images, and video contents. However, their application to NFT arts have been limited. In this paper, a GAN-based architecture is implemented and evaluated for novel NFT-style digital arts generation. Results from the qualitative case study indicate that the generated artworks are comparable to the real samples in terms of being interesting and inspiring and they were judged to be more innovative than real samples.